ISRO & Smart India Hackathon 2026·Geospatial InnovationDepthWizard
डेप्थविज़ार्डFrom Imagery to Insight
DepthWizard turns a single overhead optical image into elevation information and an interactive 3D terrain model — no stereo pair, no LiDAR pass. Upload one frame, and explore the surface it describes: hills, ridges, structures, and slope, rendered as a navigable mesh.
128×128
demo grid resolution
7 stages
reconstruction pipeline
rDSM / DSM
output modes
One photograph holds more elevation information than it looks like.
Traditional elevation mapping needs stereo imagery, LiDAR, or ground control. DepthWizard is built around a different question: how much can be recovered from a single view? The frontend below is what a geospatial analyst uses to upload imagery, track a reconstruction job, and inspect the resulting surface.
01
A single image goes in
One optical frame — a satellite tile, a drone photograph, or an aerial scan — in PNG, JPEG, TIFF, or GeoTIFF.
02
Monocular cues become height
Shading, shadow direction, texture gradients, and object scale are used to infer relative surface height across the frame.
03
A terrain model comes out
The estimated heights become a continuous elevation surface — viewable, measurable, and exportable as a mesh.
Built for reading terrain, not just viewing it
Each panel below is a working part of the workspace — not a marketing illustration.
Monocular height estimation
Recovers relative elevation from a single optical image, without a stereo pair or overlapping passes.
Contour & slope analysis
Generates contour lines and per-cell slope directly from the reconstructed surface for terrain reading.
In-scene measurement
Pick two points on the terrain to read horizontal distance, elevation difference, and 3D distance.
Elevation & slope statistics
Minimum, maximum, mean, and relief are computed automatically, with distribution charts for both.
Validation dashboards
Track demo accuracy metrics — MAE, RMSE, correlation — broken down by terrain category.
Project workspace
Every upload becomes a tracked project with its own processing history, viewer state, and analysis.
Seven stages, start to surface
The processing workspace tracks a project through each of these stages in order, with live progress for the current one.
- 01
Image upload
Transferring source imagery and verifying file integrity.
- 02
Preprocessing
Normalizing radiometry, correcting artifacts, and tiling the image.
- 03
Height estimation
Estimating relative surface height from monocular cues.
- 04
Scale calibration
Aligning relative heights against reference control where available.
- 05
Surface reconstruction
Building a continuous elevation surface from estimated heights.
- 06
Mesh generation
Converting the elevation surface into a renderable 3D mesh.
- 07
Analysis
Computing terrain statistics and preparing viewer assets.
A terrain you can actually fly through
The 3D viewer renders reconstructed terrain as a real Three.js mesh — orbit around it, switch between elevation, slope, and contour shading, drop into a first-person flythrough with WASD, or click any point to read its elevation and slope directly from the underlying height data.
- Elevation
- Slope
- Contour
- Measurement
- Flythrough