family 01
Population Household Denominator
TAM cannot be right if the denominator is wrong; this is the first reality anchor.
Validation target: district/city household and population rollups, then internal serviceable households
Prior art landscapeEditorial explainerLeakage-gated TAM
A reader for the operator who has to decide what belongs in the model
India grid TAM prediction: a prior-art reader
A standalone research map for the local prior-art corpus. It separates population, welfare, POI, road, land-mask, graph, and business-reality layers so GeoIQ-style vendor TAM remains a benchmark label rather than a production input.
read
Understand the signal stack before any repo is promoted.
gate
Keep leakage checks and city holdouts beside every metric.
map
Route each prior art to a TAM feature family and owner.
adopt
Use only versioned sources with coverage and outcome validation.
I
Chapter one - the stack
pp. 1-4
Executive Map
The practical stack is not one model. It is a controlled sequence of independent denominator, welfare, activity, serviceability, graph context, and internal outcome layers.
You are here in the mindmap
This section belongs to Map 1 - stack order.
Story role
Name the layers before choosing tools.
Carry forward
A repo is useful only after it is tied to a specific layer and validation gate.
grouped by:
independent signals denominator and welfare
context roads, POIs, graphs
gates leakage and holdouts
family 02
Buildings And Settlement Structure
Separates dense residential demand from empty land, industrial sheds, and underbuilt cells.
Validation target: audited buildings/households and installable address density
family 03
Satellite Welfare Affluence
Maps visible welfare and housing quality into the target-income/affordability band.
Validation target: Census/SECC/NFHS/SHRUG welfare indicators and internal ARPU/plan affordability
family 04
Nightlights And Economic Activity
Captures economic intensity and electrification that pure population misses.
Validation target: income/development proxies, leads, installs, ARPU, and retained revenue
family 05
Roads Accessibility Serviceability
Converts theoretical demand into reachable/serviceable demand.
Validation target: serviceability pass/fail, failed installs, partner travel time, network coverage
family 06
Poi Urban Function
Distinguishes residential <10 LPA opportunity cells from purely commercial, industrial, or sparse cells.
Validation target: lead density, conversion, install density, and local ops knowledge
family 07
Land Use Exclusions And Risk
Prevents false TAM where households cannot or should not exist.
Validation target: manual QA, failed installs, low-address-density cells, and land/water masks
family 08
Heterogeneous Graph And Spatial Context
Replaces leaky neighbor-TAM interpolation with independent neighbour/context features.
Validation target: spatial-block holdout against vendor TAM and real business outcomes
family 09
Internal Business Reality
This is the actual business reality layer; without it, confidence is only proxy confidence.
Validation target: time-based and spatial-block holdout outcomes
II
Chapter two - validity
pp. 5-8
Metric And Leakage Context
This report keeps the repository rule intact: do not present optimistic metrics without leakage checks. The useful GeoHG-style numbers are spatial-block and city holdouts; random CV remains a diagnostic warning.
You are here in the mindmap
This section belongs to Map 2 - validity gates.
Story role
Separate diagnostic metrics from defensible methodology claims.
Carry forward
Any prior-art adoption must preserve the no-vendor-TAM-feature policy.
Vendor TAM is a benchmark label for evaluation and distillation research. It is not allowed as a production score feature, neighbor aggregate, target-derived rank, or tuning signal.
manifest
Current GeoHG-style feature manifest
- Status: ok
- Vendor TAM used as feature: False
- Vendor TAM used for evaluation only: True
- WorldCover raster available: False
local figure
Generated local visual artifact
The existing scatter plot is linked below when the HTML is opened from the outputs directory.
III
Chapter three - routing
pp. 9-12
Prior-Art Signal Stack
Each family below names the prior-art repos that matter, the signal it contributes, and the validation target that must pass before it becomes more than research context.
You are here in the mindmap
This section belongs to Map 3 - feature families.
Story role
Convert repo names into auditable workstreams.
Carry forward
Every later repo section should point back to at least one family here.
| Feature family | Repos | Signals | Why it matters | Validation target |
|---|---|---|---|---|
| population household denominator | prs-eth/Popcorn, yashveeeeeeer/india-geodata, devdatalab/shrug-public, pigshell/india-census-2011 | WorldPop population, census/SHRUG reconciliation, built-up occupancy, household density | TAM cannot be right if the denominator is wrong; this is the first reality anchor. | district/city household and population rollups, then internal serviceable households |
| buildings and settlement structure | ramSeraph/indian_buildings, yashveeeeeeer/india-geodata, prs-eth/Popcorn | building_count, built_area_share, building_density, occupancy_proxy, settlement_compactness | Separates dense residential demand from empty land, industrial sheds, and underbuilt cells. | audited buildings/households and installable address density |
| satellite welfare affluence | AIandGlobalDevelopmentLab/EOML-for-India, amangupt01/Village_Development_Model, mani-shailesh/satimage, torchgeo/torchgeo | roof/material proxy, lighting proxy, drinking-water proxy, Landsat/Sentinel embeddings, village development score | Maps visible welfare and housing quality into the target-income/affordability band. | Census/SECC/NFHS/SHRUG welfare indicators and internal ARPU/plan affordability |
| nightlights and economic activity | amangupt01/Village_Development_Model, yashveeeeeeer/india-geodata, devdatalab/shrug-public | VIIRS mean, VIIRS trend, nightlight blob score, commercial activity proxy | Captures economic intensity and electrification that pure population misses. | income/development proxies, leads, installs, ARPU, and retained revenue |
| roads accessibility serviceability | kraina-ai/srai, Calychas/highway2vec, yashveeeeeeer/india-geodata, ramSeraph/indian_land_features | road_length_by_class, distance_to_major_road, road_embedding, travel_friction, non-serviceable terrain | Converts theoretical demand into reachable/serviceable demand. | serviceability pass/fail, failed installs, partner travel time, network coverage |
| poi urban function | kraina-ai/srai, kraina-ai/hex2vec, PaddlePaddle/PaddleSpatial, yashveeeeeeer/india-geodata | POI counts by category, Hex2Vec/ContextualCount embeddings, schools/healthcare/markets, urban function vector | Distinguishes residential <10 LPA opportunity cells from purely commercial, industrial, or sparse cells. | lead density, conversion, install density, and local ops knowledge |
| land use exclusions and risk | ramSeraph/indian_land_features, yashveeeeeeer/india-geodata | water_share, forest_share, mining_share, industrial_land_share, flood/coastal risk, slope/elevation | Prevents false TAM where households cannot or should not exist. | manual QA, failed installs, low-address-density cells, and land/water masks |
| heterogeneous graph and spatial context | CityMind-Lab/GeoHG, kraina-ai/srai, seai-lab/TorchSpatial, wherobots/GeoTorchAI | neighbor context, semantic similarity, land-cover hypernodes, POI hypernodes, location encoders | Replaces leaky neighbor-TAM interpolation with independent neighbour/context features. | spatial-block holdout against vendor TAM and real business outcomes |
| internal business reality | firm internal data, not GitHub | leads, installs, retained installs, gross margin, CAC, failed installs, serviceability checks, capacity | This is the actual business reality layer; without it, confidence is only proxy confidence. | time-based and spatial-block holdout outcomes |
Each repo chapter below answers what the prior art does, what it consumes, what it emits, how it maps to TAM, and what must be validated before use.
IV
Prior art 01 - Graph socioeconomic inference
pp. 13-15
GeoHG: heterogeneous graphs for socioeconomic indicators
Turns geographic raster/vector features into a heterogeneous graph with area nodes, semantic entity hypernodes, POI hypernodes, and spatial adjacency.
You are here in the mindmap
This section belongs to Map 4 - Graph socioeconomic inference.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Graph socioeconomic inference
Closest methodological fit
heterogeneous graph and spatial context
poi urban function
land use exclusions and risk
inputs
grid cells, land cover, POIs
method
Uses message passing to combine local neighborhood context with global semantic similarity across land-cover or POI categories.
outputs
grid predictions, graph embeddings, area-area edges
tam use
Use it to encode independent spatial context without using neighbor vendor TAM.
gate
vendor_tam_used_as_feature must be false
what it does
Core Behavior
- Turns geographic raster/vector features into a heterogeneous graph with area nodes, semantic entity hypernodes, POI hypernodes, and spatial adjacency.
- Uses message passing to combine local neighborhood context with global semantic similarity across land-cover or POI categories.
- Targets socioeconomic indicators such as population, GDP, night light, PM2.5, and carbon from spatial inputs.
- Provides a direct conceptual template for the local GeoHG-style feature builder already used in this repository.
model input
Inputs
- grid cells
- land cover
- POIs
- raster indicators
- observed indicator labels
model output
Outputs
- grid predictions
- graph embeddings
- area-area edges
- entity-area edges
- POI-area edges
tam role
How It Helps TAM
- Use it to encode independent spatial context without using neighbor vendor TAM.
- Preserve area-area adjacency and semantic hyperedges for soil, land use, POIs, flood risk, and nightlight bins.
- Evaluate only with spatial-block and city holdouts before any business claim.
promotion gate
Validation Gates
- vendor_tam_used_as_feature must be false
- random CV is leakage warning only when position encodings exist
- city holdout and no-position semantic runs must be reported
watch
Risks And Caveats
- Full GNN training can hide leakage if labels or GeoIQ-derived aggregates enter node features.
- Coordinates can dominate if not separated from semantic features.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
V
Prior art 02 - Population denominator
pp. 16-18
POPCORN: population from Sentinel imagery and coarse census counts
Estimates high-resolution population maps from Sentinel-1 and Sentinel-2 imagery with a small number of coarse census counts.
You are here in the mindmap
This section belongs to Map 5 - Population denominator.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Population denominator
Strong denominator challenger
population household denominator
buildings and settlement structure
inputs
Sentinel-1, Sentinel-2, coarse census totals
method
Separates built-up extraction from occupancy modeling, which helps distinguish populated residential area from empty built-up land.
outputs
population density, built-up map, occupancy proxy
tam use
Use as the population/household denominator challenger against WorldPop, Census, SHRUG, and building redistribution.
gate
district and city rollups must reconcile against official totals
what it does
Core Behavior
- Estimates high-resolution population maps from Sentinel-1 and Sentinel-2 imagery with a small number of coarse census counts.
- Separates built-up extraction from occupancy modeling, which helps distinguish populated residential area from empty built-up land.
- Produces interpretable population, built-up, and occupancy outputs that can be aggregated to 100 m, 1 km, or H3 cells.
- Includes inference, training, and data preparation paths for reproducible population mapping in data-scarce regions.
model input
Inputs
- Sentinel-1
- Sentinel-2
- coarse census totals
- boundary rasters
model output
Outputs
- population density
- built-up map
- occupancy proxy
- population change map
tam role
How It Helps TAM
- Use as the population/household denominator challenger against WorldPop, Census, SHRUG, and building redistribution.
- Treat occupancy as a residential confidence input, not as income truth.
- Aggregate predictions to H3 and vendor-compatible grid cells after source versioning.
promotion gate
Validation Gates
- district and city rollups must reconcile against official totals
- building occupancy must be audited against address/install density
- imagery windows must be pinned before evaluation
watch
Risks And Caveats
- Population is not TAM until income band, serviceability, and acquirability are estimated.
- Transfer from non-India training areas needs validation.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
VI
Prior art 03 - Geospatial feature engineering toolkit
pp. 19-21
SRAI: spatial representations for artificial intelligence
Downloads and processes OSM, Overture, GTFS, vector data, microregions, and embeddings.
You are here in the mindmap
This section belongs to Map 6 - Geospatial feature engineering toolkit.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Geospatial feature engineering toolkit
Production-friendly toolkit
poi urban function
roads accessibility serviceability
heterogeneous graph and spatial context
inputs
OSM, Overture Maps, GTFS
method
Provides H3 and other regionalization, loaders, embedders, datasets, and visualization utilities.
outputs
region feature tables, spatial embeddings, H3 vectors
tam use
Use it to standardize H3 feature construction for roads, POIs, OSM tags, and region embeddings.
gate
OSM/Overture extract date must be recorded
what it does
Core Behavior
- Downloads and processes OSM, Overture, GTFS, vector data, microregions, and embeddings.
- Provides H3 and other regionalization, loaders, embedders, datasets, and visualization utilities.
- Implements or supersedes Hex2Vec and related geospatial embedding pipelines.
- Turns raw spatial layers into reusable region vectors for downstream models.
model input
Inputs
- OSM
- Overture Maps
- GTFS
- H3 cells
- POIs
- roads
- buildings
model output
Outputs
- region feature tables
- spatial embeddings
- H3 vectors
- benchmark datasets
tam role
How It Helps TAM
- Use it to standardize H3 feature construction for roads, POIs, OSM tags, and region embeddings.
- Pair its embeddings with transparent counts so reason codes remain explainable.
- Use dated extracts rather than live OSM pulls for modeling runs.
promotion gate
Validation Gates
- OSM/Overture extract date must be recorded
- coverage null rates must be reported by city and state
- embeddings must be compared to count baselines
watch
Risks And Caveats
- OSM sparsity can look like low demand unless completeness is separately measured.
- Embeddings are hard to explain without companion feature counts.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
VII
Prior art 04 - OSM/H3 embeddings
pp. 22-24
Hex2Vec: H3 region embeddings from OSM tags
Learns context-aware embeddings for H3 hexagons from OpenStreetMap tags.
You are here in the mindmap
This section belongs to Map 7 - OSM/H3 embeddings.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
OSM/H3 embeddings
Superseded by SRAI but conceptually important
poi urban function
heterogeneous graph and spatial context
inputs
H3 cells, OSM tags, neighbor context
method
Uses neighboring regions to produce urban function vectors from map features.
outputs
hex embeddings, urban function vectors, similarity maps
tam use
Use for urban function vectors that separate residential, commercial, education, market, and sparse cells.
gate
dated OSM extract required
what it does
Core Behavior
- Learns context-aware embeddings for H3 hexagons from OpenStreetMap tags.
- Uses neighboring regions to produce urban function vectors from map features.
- Frames POI and land-use context as a representation learning problem rather than manual indicators only.
- Acts as the conceptual basis for SRAI-style region embedders.
model input
Inputs
- H3 cells
- OSM tags
- neighbor context
- regional graph
model output
Outputs
- hex embeddings
- urban function vectors
- similarity maps
tam role
How It Helps TAM
- Use for urban function vectors that separate residential, commercial, education, market, and sparse cells.
- Compare embedding clusters with manual reason-code categories.
- Use only after source completeness flags are attached.
promotion gate
Validation Gates
- dated OSM extract required
- embedding clusters must be inspected manually
- must not tune embedding choice on GeoIQ TAM correlation
watch
Risks And Caveats
- The original repo is not maintained; use SRAI for implementation.
- Sparse OSM regions can create false negatives.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
VIII
Prior art 05 - Road accessibility embeddings
pp. 25-27
Highway2Vec: road-network representations for microregions
Represents OSM microregions through road network characteristics.
You are here in the mindmap
This section belongs to Map 8 - Road accessibility embeddings.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
Road accessibility embeddings
Useful road morphology reference
roads accessibility serviceability
inputs
OSM roads, microregions, road classes
method
Builds road feature datasets and trains autoencoder-style embeddings for road morphology.
outputs
road embeddings, road-density features, microregion similarity
tam use
Use for reachability, install feasibility, service cost, partner routing, and dense-lane indicators.
gate
validate against failed installs and serviceability checks
what it does
Core Behavior
- Represents OSM microregions through road network characteristics.
- Builds road feature datasets and trains autoencoder-style embeddings for road morphology.
- Turns street structure, density, and connectivity into compact vectors.
- Offers a pathway from raw road geometries to serviceability and accessibility features.
model input
Inputs
- OSM roads
- microregions
- road classes
- network topology
model output
Outputs
- road embeddings
- road-density features
- microregion similarity
tam role
How It Helps TAM
- Use for reachability, install feasibility, service cost, partner routing, and dense-lane indicators.
- Pair with explicit road length by class and distance-to-major-road features.
- Use as a serviceable TAM component rather than gross demand.
promotion gate
Validation Gates
- validate against failed installs and serviceability checks
- compare road embeddings to transparent road counts
- document OSM road coverage per city
watch
Risks And Caveats
- Road density can mark commercial/industrial corridors, not residential TAM.
- Road network availability differs by mapping quality.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
IX
Prior art 06 - Remote-sensing ML backbone
pp. 28-30
TorchGeo: PyTorch datasets, samplers, transforms, and geospatial models
Provides PyTorch-native geospatial datasets, samplers, transforms, trainers, and pretrained models.
You are here in the mindmap
This section belongs to Map 9 - Remote-sensing ML backbone.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Remote-sensing ML backbone
Core remote-sensing toolkit
satellite welfare affluence
land use exclusions and risk
buildings and settlement structure
inputs
Landsat, Sentinel, raster labels
method
Handles geospatial metadata, CRS alignment, raster sampling, dataset intersection, and large-image patch workflows.
outputs
imagery embeddings, land-cover models, segmentation masks
tam use
Use for built-up texture, land-cover masks, settlement morphology, and satellite welfare residual features.
gate
image source manifest and CRS must be locked
what it does
Core Behavior
- Provides PyTorch-native geospatial datasets, samplers, transforms, trainers, and pretrained models.
- Handles geospatial metadata, CRS alignment, raster sampling, dataset intersection, and large-image patch workflows.
- Supports classification, regression, segmentation, detection, and change detection tasks.
- Makes satellite feature extraction repeatable when transparent tabular baselines need imagery features.
model input
Inputs
- Landsat
- Sentinel
- raster labels
- masks
- patch samplers
model output
Outputs
- imagery embeddings
- land-cover models
- segmentation masks
- remote-sensing features
tam role
How It Helps TAM
- Use for built-up texture, land-cover masks, settlement morphology, and satellite welfare residual features.
- Keep imagery models behind simpler baselines until residuals justify them.
- Archive imagery source windows and train/validation split geometry.
promotion gate
Validation Gates
- image source manifest and CRS must be locked
- spatial holdouts required for every imagery model
- compare to non-image transparent baseline
watch
Risks And Caveats
- Random patch splits can leak local visual texture.
- Pretrained embeddings may be hard to explain to business users.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
X
Prior art 07 - Scalable raster and spatiotemporal ML
pp. 31-33
GeoTorchAI: spatiotemporal deep learning with PyTorch and Sedona
Combines PyTorch models with Apache Sedona preprocessing for raster imagery and spatiotemporal non-imagery data.
You are here in the mindmap
This section belongs to Map 10 - Scalable raster and spatiotemporal ML.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 3
Scalable raster and spatiotemporal ML
Use when scale/time justifies it
satellite welfare affluence
heterogeneous graph and spatial context
inputs
raster imagery, spatiotemporal grids, Spark/Sedona tables
method
Includes datasets, transforms, models, and scalable preprocessing patterns.
outputs
trained models, spatiotemporal predictions, raster features
tam use
Use only if H3 features become time-indexed and the project needs scalable raster/time processing.
gate
time split must be enforced
what it does
Core Behavior
- Combines PyTorch models with Apache Sedona preprocessing for raster imagery and spatiotemporal non-imagery data.
- Includes datasets, transforms, models, and scalable preprocessing patterns.
- Targets satellite classification/segmentation and spatiotemporal prediction tasks.
- Fits larger data engineering workloads than a notebook-only pipeline.
model input
Inputs
- raster imagery
- spatiotemporal grids
- Spark/Sedona tables
- time series
model output
Outputs
- trained models
- spatiotemporal predictions
- raster features
tam role
How It Helps TAM
- Use only if H3 features become time-indexed and the project needs scalable raster/time processing.
- Could support lead/install time windows, nightlight trends, and seasonal imagery inputs.
- Keep simple pandas/geopandas pipeline until scale is the actual blocker.
promotion gate
Validation Gates
- time split must be enforced
- Spark/Sedona versions must be recorded
- business outcome timestamps required for temporal claims
watch
Risks And Caveats
- Heavier stack may slow auditability.
- Not India-specific and not directly a TAM method.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XI
Prior art 08 - Location representation and bias audit
pp. 34-36
TorchSpatial: location encoders and geographic bias benchmarks
Implements a framework and benchmark suite for spatial representation learning.
You are here in the mindmap
This section belongs to Map 11 - Location representation and bias audit.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
Location representation and bias audit
Bias/coordinate encoder reference
heterogeneous graph and spatial context
inputs
coordinates, location labels, geo-aware image/regression datasets
method
Includes many location encoders, LocBench datasets, and geo-bias metrics.
outputs
location embeddings, benchmark metrics, geo-bias scores
tam use
Use to compare raw coordinates, city-relative coordinates, and learned location encoders.
gate
coordinate-only baseline must be reported
what it does
Core Behavior
- Implements a framework and benchmark suite for spatial representation learning.
- Includes many location encoders, LocBench datasets, and geo-bias metrics.
- Separates coordinate encoding performance from geographic bias concerns.
- Helps test whether a model is using useful spatial priors or memorizing location.
model input
Inputs
- coordinates
- location labels
- geo-aware image/regression datasets
model output
Outputs
- location embeddings
- benchmark metrics
- geo-bias scores
tam role
How It Helps TAM
- Use to compare raw coordinates, city-relative coordinates, and learned location encoders.
- Use geo-bias framing to explain why coordinate-only baselines are not enough.
- Keep no-position semantic model alongside any coordinate-aware model.
promotion gate
Validation Gates
- coordinate-only baseline must be reported
- semantic/no-position model must be reported
- city holdout must not collapse
watch
Risks And Caveats
- Location encoders can become spatial lookup tables.
- Strong coordinate fit does not mean business causality.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XII
Prior art 09 - Urban computing toolkit
pp. 37-39
PaddleSpatial: spatiotemporal computing and region profiling
Provides spatial-temporal data mining functions for transfer learning, time-series prediction, and region profiling.
You are here in the mindmap
This section belongs to Map 12 - Urban computing toolkit.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 3
Urban computing toolkit
Conceptual reference
poi urban function
heterogeneous graph and spatial context
source custody
Local path: prior art/PaddlePaddle_PaddleSpatial
Upstream: PaddlePaddle/PaddleSpatial
inputs
urban regions, spatiotemporal signals, POIs
method
Frames urban computing applications around reusable spatial and temporal feature modules.
outputs
region profiles, time-series predictions, urban computing features
tam use
Use as a reference for region profile modules and spatiotemporal abstractions.
gate
needs internal time-indexed outcomes
what it does
Core Behavior
- Provides spatial-temporal data mining functions for transfer learning, time-series prediction, and region profiling.
- Frames urban computing applications around reusable spatial and temporal feature modules.
- Includes tutorials and API surfaces for broader region modeling.
- Useful as a reference for region profiling, not as an India-specific data source.
model input
Inputs
- urban regions
- spatiotemporal signals
- POIs
- mobility or demand series
model output
Outputs
- region profiles
- time-series predictions
- urban computing features
tam role
How It Helps TAM
- Use as a reference for region profile modules and spatiotemporal abstractions.
- Consider if internal outcomes become dense time series.
- Do not treat it as a source of India geography.
promotion gate
Validation Gates
- needs internal time-indexed outcomes
- must be compared with simpler region profiles
- source coverage must be independent of GeoIQ
watch
Risks And Caveats
- Not tailored to the current India data stack.
- May introduce framework complexity before needed.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XIII
Prior art 10 - India-specific welfare modeling
pp. 40-42
EOML-for-India: satellite imagery for Indian health and living standards
Reproduces research on measuring Indian living standards and health indicators from satellite images.
You are here in the mindmap
This section belongs to Map 13 - India-specific welfare modeling.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
India-specific welfare modeling
Strong India-specific welfare prior
satellite welfare affluence
population household denominator
source custody
Local path: prior art/AIandGlobalDevelopmentLab_EOML-for-India
Upstream: AIandGlobalDevelopmentLab/EOML-for-India
inputs
Indian village coordinates, Landsat imagery, Census labels
method
Downloads imagery, retrieves village image chips, cleans census labels, and trains models.
outputs
living-standard predictions, health proxy predictions, village image crops
tam use
Use as the India-specific welfare/affluence modeling template for income-band probability.
gate
village-to-cell crosswalk quality required
what it does
Core Behavior
- Reproduces research on measuring Indian living standards and health indicators from satellite images.
- Downloads imagery, retrieves village image chips, cleans census labels, and trains models.
- Uses Indian village geography and census-style indicators as welfare supervision.
- Provides a strong template for satellite-to-welfare modeling in India.
model input
Inputs
- Indian village coordinates
- Landsat imagery
- Census labels
- NFHS-style indicators
model output
Outputs
- living-standard predictions
- health proxy predictions
- village image crops
tam role
How It Helps TAM
- Use as the India-specific welfare/affluence modeling template for income-band probability.
- Map welfare outputs to H3 cells only after reconciling village/cell geometry.
- Use as proxy confidence, not direct household income truth.
promotion gate
Validation Gates
- village-to-cell crosswalk quality required
- temporal split or city/state holdout needed
- welfare target definitions must be documented
watch
Risks And Caveats
- Village labels may not transfer directly to dense urban broadband TAM.
- Historical census targets can be stale without calibration.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XIV
Prior art 11 - India-specific rural and peri-urban development
pp. 43-45
Village Development Model: Indian village imagery, nightlights, and population features
Downloads Indian state imagery through Google Earth Engine and cuts it into fixed village image crops.
You are here in the mindmap
This section belongs to Map 14 - India-specific rural and peri-urban development.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
India-specific rural and peri-urban development
Operational template for village-scale work
satellite welfare affluence
nightlights and economic activity
population household denominator
source custody
Local path: prior art/amangupt01_Village_Development_Model
Upstream: amangupt01/Village_Development_Model
inputs
GEE imagery, village boundaries, census indicators
method
Builds census-derived development levels, population features, nightlight features, and nearest-neighbor features.
outputs
development scores, nightlight features, population features
tam use
Use for rural/peri-urban welfare proxy design and validation patterns.
gate
village imagery cloud and exclusion report required
what it does
Core Behavior
- Downloads Indian state imagery through Google Earth Engine and cuts it into fixed village image crops.
- Builds census-derived development levels, population features, nightlight features, and nearest-neighbor features.
- Trains CNN and regression architectures for development indicators.
- Includes hypothesis testing, visualizations, occlusion studies, error analysis, and statewise statistics.
model input
Inputs
- GEE imagery
- village boundaries
- census indicators
- VIIRS/nightlights
- population features
model output
Outputs
- development scores
- nightlight features
- population features
- error analysis maps
tam role
How It Helps TAM
- Use for rural/peri-urban welfare proxy design and validation patterns.
- Borrow its nightlight and population feature engineering before training heavy imagery models.
- Use occlusion/error analysis style for model interpretability.
promotion gate
Validation Gates
- village imagery cloud and exclusion report required
- spatial/state holdout required
- development labels must not be used as current income truth without calibration
watch
Risks And Caveats
- Village-scale development does not directly equal urban 0-10 LPA TAM.
- GEE imagery exports need reproducible date windows.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XV
Prior art 12 - Interpretable satellite welfare proxies
pp. 46-48
satimage: poverty prediction from roof, lighting, water, and satellite imagery
Predicts developmental parameters from satellite images using a multi-task fully convolutional model.
You are here in the mindmap
This section belongs to Map 15 - Interpretable satellite welfare proxies.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Interpretable satellite welfare proxies
Interpretable housing-quality prior
satellite welfare affluence
buildings and settlement structure
inputs
satellite image chips, Census 2011, SECC income labels
method
Targets roof material, lighting source, and drinking-water source as interpretable welfare proxies.
outputs
roof probabilities, lighting probabilities, water-source probabilities
tam use
Use outputs as reason-code-friendly welfare signals for target-income probability.
gate
manual image QA for representative cells
what it does
Core Behavior
- Predicts developmental parameters from satellite images using a multi-task fully convolutional model.
- Targets roof material, lighting source, and drinking-water source as interpretable welfare proxies.
- Uses predicted or actual developmental parameters to estimate income/poverty levels.
- Includes model weights, sample data, filter responses, and learning-curve materials.
model input
Inputs
- satellite image chips
- Census 2011
- SECC income labels
- region coordinates
model output
Outputs
- roof probabilities
- lighting probabilities
- water-source probabilities
- income/poverty predictions
tam role
How It Helps TAM
- Use outputs as reason-code-friendly welfare signals for target-income probability.
- Translate roof/lighting/water predictions into interpretable affluence bands.
- Pair with current nightlight and building features for freshness.
promotion gate
Validation Gates
- manual image QA for representative cells
- spatial holdout by region
- calibrate old census/SECC targets against internal ARPU/conversion
watch
Risks And Caveats
- Satellite visual proxies can encode rural housing better than apartment-heavy urban density.
- Static Maps inputs and old labels need modern replacement.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XVI
Prior art 13 - India-specific academic reference
pp. 49-51
Socio-economic indicators using satellite imagery
Contains notebooks and thesis material for satellite-based socioeconomic indicator estimation.
You are here in the mindmap
This section belongs to Map 16 - India-specific academic reference.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 4
India-specific academic reference
Optional context
satellite welfare affluence
source custody
Local path: prior art/ArmaanBhullar_Socio-Economic-Indicators-using-Satellite-Imagery
Upstream: ArmaanBhullar/Socio-Economic-Indicators-using-Satellite-Imagery
inputs
selected villages, satellite-derived features, socioeconomic indicators
method
Includes village selection, data analysis, PCA, regression, and feature notes.
outputs
analysis notebooks, PCA outputs, regression outputs
tam use
Use as optional background for indicator engineering and notebook narrative.
gate
notebook cell length standards apply before edits
what it does
Core Behavior
- Contains notebooks and thesis material for satellite-based socioeconomic indicator estimation.
- Includes village selection, data analysis, PCA, regression, and feature notes.
- Acts as a lightweight India-specific academic reference rather than a production toolkit.
- Can help compare explanatory notebook style and feature lists.
model input
Inputs
- selected villages
- satellite-derived features
- socioeconomic indicators
model output
Outputs
- analysis notebooks
- PCA outputs
- regression outputs
tam role
How It Helps TAM
- Use as optional background for indicator engineering and notebook narrative.
- Do not prioritize over EOML, Village Development Model, or satimage.
- Mine only if it contains a useful feature or validation idea.
promotion gate
Validation Gates
- notebook cell length standards apply before edits
- features must be reconciled to current source manifests
- metrics must not be copied without leakage review
watch
Risks And Caveats
- Sparse repository and limited maintenance.
- Likely not enough for operational India-wide scoring.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XVII
Prior art 14 - India socioeconomic geography backbone
pp. 52-54
SHRUG: Socioeconomic High-resolution Rural-Urban Geographic Platform for India
Tracks releases and issues for SHRUG, the India rural-urban socioeconomic geography platform.
You are here in the mindmap
This section belongs to Map 17 - India socioeconomic geography backbone.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
India socioeconomic geography backbone
Must-use backbone
population household denominator
satellite welfare affluence
nightlights and economic activity
inputs
SHRUG releases, village/town polygons, socioeconomic tables
method
Connects villages, towns, administrative geography, socioeconomic fields, environmental data, and nightlights.
outputs
SHRID geography, socioeconomic joins, rural/urban indicators
tam use
Use for village/town socioeconomic reconciliation, denominator checks, and geographic crosswalks.
gate
release version and license must be documented
what it does
Core Behavior
- Tracks releases and issues for SHRUG, the India rural-urban socioeconomic geography platform.
- Connects villages, towns, administrative geography, socioeconomic fields, environmental data, and nightlights.
- Provides high-resolution India geography and socioeconomic context outside vendor TAM.
- Acts as a reconciliation backbone for Census, boundaries, villages, towns, and indicators.
model input
Inputs
- SHRUG releases
- village/town polygons
- socioeconomic tables
- nightlights
model output
Outputs
- SHRID geography
- socioeconomic joins
- rural/urban indicators
- release metadata
tam role
How It Helps TAM
- Use for village/town socioeconomic reconciliation, denominator checks, and geographic crosswalks.
- Use SHRUG fields as independent welfare/context features after licensing and versioning.
- Use for city/rural holdout segmentation and geographic rollups.
promotion gate
Validation Gates
- release version and license must be documented
- cell-to-SHRID join success required
- fields must be dated and not target-derived
watch
Risks And Caveats
- Release data may sit outside the GitHub issue repo.
- Some licensing terms may restrict commercial use.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XVIII
Prior art 15 - India geodata aggregator
pp. 55-57
India Geodata: unified open India geospatial data collection
Aggregates India administrative, census, environment, water, infrastructure, buildings, healthcare, education, urban, postal, police, and remote-sensing layers.
You are here in the mindmap
This section belongs to Map 18 - India geodata aggregator.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
India geodata aggregator
Local feature backbone
population household denominator
roads accessibility serviceability
poi urban function
land use exclusions and risk
nightlights and economic activity
source custody
Local path: prior art/yashveeeeeeer_india-geodata
Upstream: yashveeeeeeer/india-geodata
inputs
government geodata, open community datasets, remote-sensing panels
method
Provides README and metadata files for many data directories.
outputs
GeoJSON, Parquet, PMTiles
tam use
Use as the first local source registry for independent features.
gate
each layer needs its own source status
what it does
Core Behavior
- Aggregates India administrative, census, environment, water, infrastructure, buildings, healthcare, education, urban, postal, police, and remote-sensing layers.
- Provides README and metadata files for many data directories.
- Contains local layers already consumed by the GeoHG-style feature builder.
- Offers immediate local sources for POIs, nightlights, flood, soil, roads, boundaries, and slum overlaps.
model input
Inputs
- government geodata
- open community datasets
- remote-sensing panels
- curated metadata
model output
Outputs
- GeoJSON
- Parquet
- PMTiles
- CSV
- shapefiles
- metadata
tam role
How It Helps TAM
- Use as the first local source registry for independent features.
- Promote each layer separately only after coverage, license, CRS, and null-rate checks.
- Use metadata and README files to keep feature provenance explicit.
promotion gate
Validation Gates
- each layer needs its own source status
- worldcover gap must remain marked missing until filled
- coverage by vendor cities and national AOI must be reported
watch
Risks And Caveats
- Aggregator convenience can hide upstream source differences.
- Large files and release assets may not all be local.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XIX
Prior art 16 - Boundary backbone
pp. 58-60
Indian administrative boundaries
Publishes Indian administrative boundary releases for states, districts, subdistricts, blocks, panchayats, villages, habitations, urban areas, forests, postal, police, constituencies, and historical districts.
You are here in the mindmap
This section belongs to Map 19 - Boundary backbone.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Boundary backbone
Critical join infrastructure
population household denominator
land use exclusions and risk
source custody
Local path: prior art/ramSeraph_indian_admin_boundaries
Upstream: ramSeraph/indian_admin_boundaries
inputs
LGD, SOI, Bhuvan
method
Emphasizes Survey of India guidance around political boundary standards.
outputs
boundary layers, admin crosswalk inputs, release artifacts
tam use
Use to attach cells to state, district, subdistrict, village, panchayat, urban, postal, and police units.
gate
join success per admin level required
what it does
Core Behavior
- Publishes Indian administrative boundary releases for states, districts, subdistricts, blocks, panchayats, villages, habitations, urban areas, forests, postal, police, constituencies, and historical districts.
- Emphasizes Survey of India guidance around political boundary standards.
- Provides a multi-level spatial backbone for joining grid cells to official geography.
- Connects naturally to indianopenmaps for tile deployment.
model input
Inputs
- LGD
- SOI
- Bhuvan
- PMGSY
- FSI
- NCSCM
- release assets
model output
Outputs
- boundary layers
- admin crosswalk inputs
- release artifacts
tam role
How It Helps TAM
- Use to attach cells to state, district, subdistrict, village, panchayat, urban, postal, and police units.
- Use as a cross-check against india-geodata and SHRUG boundaries.
- Record unresolved or conflicting cells as confidence penalties.
promotion gate
Validation Gates
- join success per admin level required
- boundary source version required
- conflict report required before rollups
watch
Risks And Caveats
- Boundary systems can disagree; do not silently merge.
- Official-looking boundaries can still be outdated relative to current admin changes.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XX
Prior art 17 - Building footprints
pp. 61-63
Indian buildings: urban, Google Open Buildings, and Microsoft Buildings
Collects Indian building footprint releases from urban sources, Google Open Buildings 2023, and Microsoft Buildings.
You are here in the mindmap
This section belongs to Map 20 - Building footprints.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Building footprints
Core physical denominator source
buildings and settlement structure
population household denominator
source custody
Local path: prior art/ramSeraph_indian_buildings
Upstream: ramSeraph/indian_buildings
inputs
urban building footprints, Google Open Buildings, Microsoft Buildings
method
Provides a physical settlement and structure layer independent of vendor TAM.
outputs
building footprints, built area features, building-count features
tam use
Use for household redistribution and residential confidence after source disagreement checks.
gate
coverage and confidence threshold by source required
what it does
Core Behavior
- Collects Indian building footprint releases from urban sources, Google Open Buildings 2023, and Microsoft Buildings.
- Provides a physical settlement and structure layer independent of vendor TAM.
- Supports building count, built area, compactness, and occupancy proxy features.
- Helps identify empty land, industrial sheds, and high-density residential pockets.
model input
Inputs
- urban building footprints
- Google Open Buildings
- Microsoft Buildings
model output
Outputs
- building footprints
- built area features
- building-count features
tam role
How It Helps TAM
- Use for household redistribution and residential confidence after source disagreement checks.
- Compare Google and Microsoft disagreement cells for confidence scoring.
- Never treat footprint count as household count without occupancy calibration.
promotion gate
Validation Gates
- coverage and confidence threshold by source required
- vertical-density caveat required in dense cities
- validate against installable address density
watch
Risks And Caveats
- High-rise households are underrepresented by footprint counts.
- Industrial/commercial footprints can inflate residential confidence.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXI
Prior art 18 - Land masks and physical constraints
pp. 64-66
Indian land features: land use, soil, forests, mining, elevation, degradation
Publishes Indian land-use, urban land-use, soil health, geomorphology, geology, forests, mining, groundwater prospects, elevation, and land degradation layers.
You are here in the mindmap
This section belongs to Map 21 - Land masks and physical constraints.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
Land masks and physical constraints
False-positive suppression source
land use exclusions and risk
roads accessibility serviceability
source custody
Local path: prior art/ramSeraph_indian_land_features
Upstream: ramSeraph/indian_land_features
inputs
land-use releases, soil, forests
method
Provides physical and environmental constraints for cell scoring.
outputs
mask layers, risk flags, terrain features
tam use
Use for land/water/forest/mining/terrain penalties and non-residential suppression.
gate
mask coverage and date required
what it does
Core Behavior
- Publishes Indian land-use, urban land-use, soil health, geomorphology, geology, forests, mining, groundwater prospects, elevation, and land degradation layers.
- Provides physical and environmental constraints for cell scoring.
- Helps separate residential opportunity from forest, mining, water-adjacent, industrial, steep, or degraded land.
- Complements land-cover rasters with India-specific thematic data.
model input
Inputs
- land-use releases
- soil
- forests
- mining
- CartoDEM
- SOI elevation
model output
Outputs
- mask layers
- risk flags
- terrain features
- land viability features
tam role
How It Helps TAM
- Use for land/water/forest/mining/terrain penalties and non-residential suppression.
- Build explicit negative reason codes for low-confidence cells.
- Keep masking effects visible so analysts can override with field evidence.
promotion gate
Validation Gates
- mask coverage and date required
- manual QA on suppressed high-demand cells
- failed-install and serviceability validation preferred
watch
Risks And Caveats
- Land-use layers may be stale or coarse.
- Over-aggressive masks can remove real dense settlements.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXII
Prior art 19 - Map serving and local QA
pp. 67-69
Indian Open Maps: tiles, PMTiles, COGs, and extraction workflows
Serves India vector/raster tiles from PMTiles and cloud-optimized geospatial files.
You are here in the mindmap
This section belongs to Map 22 - Map serving and local QA.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
Map serving and local QA
Useful for visual QA tooling
land use exclusions and risk
poi urban function
inputs
PMTiles, COGs, boundary releases
method
Includes Python and processing utilities for extraction and filtering.
outputs
map tiles, extracts, visual QA layers
tam use
Use to make grid, feature, mask, and residual layers inspectable by operations teams.
gate
tile build metadata required
what it does
Core Behavior
- Serves India vector/raster tiles from PMTiles and cloud-optimized geospatial files.
- Includes Python and processing utilities for extraction and filtering.
- Turns large boundary/source layers into web map layers for QA and exploration.
- Supports a local visual explorer for grids, masks, and feature disagreement.
model input
Inputs
- PMTiles
- COGs
- boundary releases
- feature layers
model output
Outputs
- map tiles
- extracts
- visual QA layers
tam role
How It Helps TAM
- Use to make grid, feature, mask, and residual layers inspectable by operations teams.
- Prefer visual QA for source coverage and suspicious high/low cells.
- Do not treat map-serving machinery as a model signal.
promotion gate
Validation Gates
- tile build metadata required
- visual layers must link to source versions
- QA maps should show uncertainty/confidence
watch
Risks And Caveats
- Visualization can look authoritative even when source coverage is partial.
- Tile generation is not a substitute for numeric validation.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXIII
Prior art 20 - City ward and municipal overlays
pp. 70-72
DataMeet Municipal Spatial Data
Collects municipal spatial data scraped from city websites and other sources.
You are here in the mindmap
This section belongs to Map 23 - City ward and municipal overlays.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 1
City ward and municipal overlays
Important urban overlay
population household denominator
poi urban function
source custody
Local path: prior art/datameet_Municipal_Spatial_Data
Upstream: datameet/Municipal_Spatial_Data
inputs
municipality websites, KML, GeoJSON
method
Provides KML and GeoJSON municipal ward and city layers for many Indian cities.
outputs
municipal ward polygons, city administrative overlays
tam use
Use for city rollups, ward-level planning, and field-ops overlays where available.
gate
city-specific source attribution required
what it does
Core Behavior
- Collects municipal spatial data scraped from city websites and other sources.
- Provides KML and GeoJSON municipal ward and city layers for many Indian cities.
- Documents that these are usually municipal wards, not Census wards.
- Gives city operations overlays for aggregation, QA, and local planning.
model input
Inputs
- municipality websites
- KML
- GeoJSON
- city ward data
model output
Outputs
- municipal ward polygons
- city administrative overlays
tam role
How It Helps TAM
- Use for city rollups, ward-level planning, and field-ops overlays where available.
- Mark municipal ward versus census ward distinctions explicitly.
- Attach ward coverage as a confidence attribute.
promotion gate
Validation Gates
- city-specific source attribution required
- ward type must be labeled
- coverage gaps must be visible in reports
watch
Risks And Caveats
- Coverage varies by city.
- Municipal wards are not always stable or comparable across sources.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXIV
Prior art 21 - Village boundary backup
pp. 73-75
DataMeet Indian village boundaries
Collects community-contributed Indian village boundaries for selected states.
You are here in the mindmap
This section belongs to Map 24 - Village boundary backup.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 3
Village boundary backup
Historical/community backup
population household denominator
source custody
Local path: prior art/datameet_indian_village_boundaries
Upstream: datameet/indian_village_boundaries
inputs
community boundary files, state folders, docs
method
Provides historical open village boundary material and a website/docs structure.
outputs
village polygons, state-specific boundary layers
tam use
Use only as a fallback boundary source or QA comparison.
gate
compare against SHRUG/ramSeraph first
what it does
Core Behavior
- Collects community-contributed Indian village boundaries for selected states.
- Provides historical open village boundary material and a website/docs structure.
- Can help compare or backfill village geography when primary sources are missing.
- Is less preferred than SHRUG, india-geodata, or ramSeraph for serious modeling.
model input
Inputs
- community boundary files
- state folders
- docs
model output
Outputs
- village polygons
- state-specific boundary layers
tam role
How It Helps TAM
- Use only as a fallback boundary source or QA comparison.
- Never silently replace a primary source with this layer.
- Record source priority and unresolved-cell count.
promotion gate
Validation Gates
- compare against SHRUG/ramSeraph first
- manual QA for selected states
- source age and coverage required
watch
Risks And Caveats
- Coverage and accuracy vary widely.
- Can introduce join inconsistencies if mixed without metadata.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXV
Prior art 22 - Census denominator reference
pp. 76-78
India Census 2011 PCA and houselisting mirror
Contains scraped Census 2011 PCA and houselisting-style tables.
You are here in the mindmap
This section belongs to Map 25 - Census denominator reference.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
Census denominator reference
Useful reproducibility mirror
population household denominator
satellite welfare affluence
source custody
Local path: prior art/pigshell_india-census-2011
Upstream: pigshell/india-census-2011
inputs
Census 2011 PCA, houselisting tables, scraped records
method
Provides a reproducible mirror for population and housing statistics.
outputs
CSV/table extracts, population and housing fields
tam use
Use for quick denominator reconciliation and housing feature probes.
gate
official-source comparison required
what it does
Core Behavior
- Contains scraped Census 2011 PCA and houselisting-style tables.
- Provides a reproducible mirror for population and housing statistics.
- Can be used for denominator QA, household/housing variables, and district rollups.
- Should be secondary to official Census tables when available.
model input
Inputs
- Census 2011 PCA
- houselisting tables
- scraped records
model output
Outputs
- CSV/table extracts
- population and housing fields
tam role
How It Helps TAM
- Use for quick denominator reconciliation and housing feature probes.
- Prefer official Census source for final source registry.
- Document age and projection/calibration method before current TAM use.
promotion gate
Validation Gates
- official-source comparison required
- join-key audit required
- staleness note required
watch
Risks And Caveats
- Scraped data may contain errors.
- 2011 values are stale for current city growth.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXVI
Prior art 23 - Postal overlay
pp. 79-81
DataMeet pincode boundaries
Provides GeoJSON pincode boundary extents for several major Indian cities.
You are here in the mindmap
This section belongs to Map 26 - Postal overlay.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 2
Postal overlay
Address QA overlay
roads accessibility serviceability
poi urban function
inputs
pincode polygons, city folders, GeoJSON
method
Helps connect grid cells to address and postal operating workflows.
outputs
postal overlays, city pincode boundaries
tam use
Use for pincode-level QA, address mismatch checks, and operational summaries.
gate
city coverage required
what it does
Core Behavior
- Provides GeoJSON pincode boundary extents for several major Indian cities.
- Helps connect grid cells to address and postal operating workflows.
- Works as a practical overlay for geocoding QA, field planning, and pincode summaries.
- Should not be treated as perfect official postal ground truth.
model input
Inputs
- pincode polygons
- city folders
- GeoJSON
model output
Outputs
- postal overlays
- city pincode boundaries
tam role
How It Helps TAM
- Use for pincode-level QA, address mismatch checks, and operational summaries.
- Keep pincode-derived features as routing/ops metadata, not income labels.
- Compare to other postal sources where important.
promotion gate
Validation Gates
- city coverage required
- postal source caveat required
- manual QA for disputed boundaries
watch
Risks And Caveats
- Pincode polygons are approximate.
- Coverage is limited to selected cities.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXVII
Prior art 24 - Postal shape construction
pp. 82-84
Pincode shapes: editable India pincode GeoJSON workflow
Provides a workflow for seeding, editing, and downloading India pincode boundaries as GeoJSON.
You are here in the mindmap
This section belongs to Map 27 - Postal shape construction.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 3
Postal shape construction
Useful QA concept
roads accessibility serviceability
inputs
pincode points, editable shapes, GeoJSON
method
Frames pincode coverage as an editable geospatial data problem.
outputs
pincode boundary candidates, downloadable shapes
tam use
Use as a workflow reference for correcting postal overlays.
gate
operator-edited shapes need provenance
what it does
Core Behavior
- Provides a workflow for seeding, editing, and downloading India pincode boundaries as GeoJSON.
- Frames pincode coverage as an editable geospatial data problem.
- Can help with address/pincode QA workflows and operational map correction.
- Is not a definitive socioeconomic or household source.
model input
Inputs
- pincode points
- editable shapes
- GeoJSON
model output
Outputs
- pincode boundary candidates
- downloadable shapes
tam role
How It Helps TAM
- Use as a workflow reference for correcting postal overlays.
- Use in QA dashboards where teams need to inspect geocoding issues.
- Do not use pincode geometry as a demand predictor by itself.
promotion gate
Validation Gates
- operator-edited shapes need provenance
- date and editor attribution required
- must be separated from model labels
watch
Risks And Caveats
- Editable shapes can drift without governance.
- Pincode boundaries are operational, not demographic truth.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXVIII
Prior art 25 - Map layer backup
pp. 85-87
Hindustan Times Labs shapefiles
Contains India, state, city, and other shapefile/GeoJSON/KML resources.
You are here in the mindmap
This section belongs to Map 28 - Map layer backup.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 4
Map layer backup
Visual/reference backup
land use exclusions and risk
source custody
Local path: prior art/HindustanTimesLabs_shapefiles
Upstream: HindustanTimesLabs/shapefiles
inputs
shapefiles, GeoJSON, KML
method
Useful for quick maps and visual context when better sources are absent.
outputs
visual map layers, backup boundaries
tam use
Use only for exploratory visualization or missing-map backup.
gate
accuracy caveat required
what it does
Core Behavior
- Contains India, state, city, and other shapefile/GeoJSON/KML resources.
- Useful for quick maps and visual context when better sources are absent.
- Carries source notes and accuracy caveats.
- Should remain backup/reference material for this project.
model input
Inputs
- shapefiles
- GeoJSON
- KML
- state/city folders
model output
Outputs
- visual map layers
- backup boundaries
tam role
How It Helps TAM
- Use only for exploratory visualization or missing-map backup.
- Prefer SHRUG, india-geodata, and ramSeraph for production joins.
- Make caveats visible in any report that includes it.
promotion gate
Validation Gates
- accuracy caveat required
- do not mix with official boundaries silently
- manual QA required for planning use
watch
Risks And Caveats
- Accuracy not guaranteed.
- Layer age and source can vary.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXIX
Prior art 26 - Legacy boundary quickstart
pp. 88-90
geohacker/india: simple India GeoJSON boundaries
Provides simple GeoJSON boundaries for Indian states, districts, and taluks.
You are here in the mindmap
This section belongs to Map 29 - Legacy boundary quickstart.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 4
Legacy boundary quickstart
Prototype-only backup
population household denominator
inputs
GADM-derived boundaries, GeoJSON
method
Useful for quick prototypes and lightweight map examples.
outputs
state boundaries, district boundaries, taluk boundaries
tam use
Use only for rough visual prototypes.
gate
prototype label required
what it does
Core Behavior
- Provides simple GeoJSON boundaries for Indian states, districts, and taluks.
- Useful for quick prototypes and lightweight map examples.
- Does not provide the depth needed for serious grid-level TAM modeling.
- Can be used as a sanity-check visual layer only.
model input
Inputs
- GADM-derived boundaries
- GeoJSON
model output
Outputs
- state boundaries
- district boundaries
- taluk boundaries
tam role
How It Helps TAM
- Use only for rough visual prototypes.
- Avoid production joins or official rollups from this source.
- Replace with current official or curated boundaries before modeling.
promotion gate
Validation Gates
- prototype label required
- source caveat required
- do not use for metrics rollup
watch
Risks And Caveats
- Legacy and likely stale.
- Insufficient for city/local operational decisions.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXX
Prior art 27 - Boundary extraction workflow
pp. 91-93
OS-Bhugol: converting government map documents into open geodata
Attempts to convert government ward/boundary PDFs and images into clean GeoJSON/KML data.
You are here in the mindmap
This section belongs to Map 30 - Boundary extraction workflow.
Story role
Decide whether this repo supplies a production input, a challenger, or context only.
Carry forward
Promote only the outputs that survive source versioning, leakage checks, and outcome validation.
classification
priority 3
Boundary extraction workflow
Watchlist workflow
poi urban function
population household denominator
inputs
government PDFs, JPEG maps, manual/geometric extraction
method
Frames hard-to-use Indian boundary documents as open spatial data extraction tasks.
outputs
GeoJSON, KML, site documentation
tam use
Use as a workflow reference for gap-filling city ward layers.
gate
document source path required
what it does
Core Behavior
- Attempts to convert government ward/boundary PDFs and images into clean GeoJSON/KML data.
- Frames hard-to-use Indian boundary documents as open spatial data extraction tasks.
- Can be useful when city wards exist only as scanned maps or PDFs.
- Current coverage appears narrow, so it is a workflow reference more than a base source.
model input
Inputs
- government PDFs
- JPEG maps
- manual/geometric extraction
- scripts
model output
Outputs
- GeoJSON
- KML
- site documentation
tam role
How It Helps TAM
- Use as a workflow reference for gap-filling city ward layers.
- Apply only with strict manual QA and source attribution.
- Store extracted boundaries as separate dated artifacts.
promotion gate
Validation Gates
- document source path required
- digitization QA required
- do not mix with official layers without conflict report
watch
Risks And Caveats
- Coverage is limited.
- Digitized boundaries can contain geometric errors.
fit bars
Feature-Family Fit
population household denominator
buildings and settlement structure
satellite welfare affluence
nightlights and economic activity
roads accessibility serviceability
poi urban function
land use exclusions and risk
heterogeneous graph and spatial context
internal business reality
XXXI
Appendix - upstream sources
pp. 94-99
Static Upstream Source Appendix
These stable upstream sources and tooling references are not automatically model-ready; each needs version, extract date, license, CRS, coverage, and leakage-safe validation before promotion.
You are here in the mindmap
This section belongs to Map 31 - upstream registry.
Story role
Separate source discovery from production readiness.
Carry forward
Treat every source as a candidate until custody and coverage checks pass.
GHSL
Independent population, built-up, and settlement benchmark.
Official Census India PCA
Official population, household, housing, and rural/urban reconciliation.
Local Government Directory
Admin-code crosswalks across Census, SHRUG, boundaries, and internal systems.
ESA WorldCover
Built-up, water, vegetation, crop, and non-residential masks.
Dynamic World V1
Time-windowed land-cover probabilities for built-up and exclusion masks.
Google Open Buildings
Building count, built area, settlement morphology, and residential-density evidence.
Microsoft Global ML Buildings
Independent building-footprint source for denominator redistribution and QA.
Overture Maps
Versioned places, buildings, roads, addresses, and base features.
OSMnx and frozen OSM extracts
Roads, POIs, network topology, and accessibility features.
ohsome API
OSM history and completeness statistics so sparse mapping is not mistaken for sparse demand.
Ookla Open Data
Quarterly broadband/mobile performance tiles as weak connectivity context.
M-Lab
Broadband measurement aggregates for independent connectivity residual analysis.
OpenCelliD
Cell tower locations as optional mobile infrastructure proxy.
Planetary Computer STAC
Reproducible raster catalog item IDs and query manifests.
DuckDB Spatial
Fast local spatial SQL for feature tables, probes, and crosswalks.
GeoParquet
Columnar spatial artifacts with CRS and metadata.
PMTiles
Compact distribution of QA map layers without a tile server.
IBM/NASA Prithvi
Foundation model embeddings for land cover and environmental context.
Clay Foundation
General satellite embeddings for welfare or settlement morphology residuals.
AllenAI SatlasPretrain
Remote-sensing pretrained models for buildings, roads, land use, and visual priors.
TerraTorch
Training and evaluation framework for geospatial foundation models.
XXXII
Appendix - adoption order
pp. 100-103
Adoption Matrix
Use this as the implementation order. It keeps the business formula auditable and avoids jumping straight from prior art to a black-box TAM score.
You are here in the mindmap
This section belongs to Map 32 - implementation order.
Story role
Turn prior-art reading into a build sequence.
Carry forward
Start with denominator and custody before adding embeddings or graph context.
| Step | Layer | Prior art | Score field | Gate |
|---|---|---|---|---|
| 1 | Denominator | WorldPop, Census, SHRUG, POPCORN, buildings | households_est | district/city rollups |
| 2 | Residential confidence | buildings, land use, WorldCover, land masks | residential_confidence | manual QA plus failed-install checks |
| 3 | Income/affluence proxy | EOML, Village Development, satimage, nightlights | income_band_prob | Census/SHRUG/NFHS/internal ARPU |
| 4 | Serviceability | roads, terrain, partner/network, failed installs | serviceable_prob | ops serviceability outcomes |
| 5 | Acquirability | leads, installs, conversion, competition, channel coverage | acquirable_prob | time and spatial holdout |
| 6 | Graph context | GeoHG, SRAI, Hex2Vec, Highway2Vec | context features | spatial-block and city holdout |
XXXIII
Appendix - repeated visual gates
pp. 104-118
Visual Evidence Atlas
This dense visual appendix maps each prior-art item to feature families and repeats the gates that keep the work audit-safe.
You are here in the mindmap
This section belongs to Map 33 - repeated gate tiles.
Story role
Make the long tail scannable without flattening it into prose.
Carry forward
Read the tiles as prompts for validation, not as readiness scores.
XXXIV
Appendix - vocabulary
pp. 119-121
Glossary
The floating lingo panel is convenient on screen; this chapter keeps the same vocabulary available when the report is printed or archived.
You are here in the mindmap
This section belongs to Map 34 - vocabulary.
Story role
Keep the report's repeated terms unambiguous.
Carry forward
When a model artifact uses one of these terms, it should use the same meaning.
| Term | Plain-English meaning |
|---|---|
| Vendor TAM | A benchmark or evaluation label only; never a production feature. |
| Spatial holdout | A validation split that withholds geographic areas, not random rows. |
| City holdout | A validation split that tests whether the model survives unseen cities. |
| Denominator | The population, household, or building base before ability-to-buy adjustments. |
| Residential confidence | Evidence that a cell contains usable residential demand rather than empty built-up area. |
| Serviceability | Whether the business can feasibly reach or install in the cell. |
| Acquirability | Whether demand can be won through channels, price, and competition. |
| OSM completeness | A guardrail that separates missing map data from missing demand. |
| GeoHG | A heterogeneous graph prior art for area, semantic, and POI context. |
| Distillation | A research step that approximates benchmark labels without allowing those labels into production features. |