Skip to content
SIH Buddyby Ganeev Singh
Dev
All problem statements
SIH26158High risk high rewardacceptance 3/5

Single-Pass Drone Video to Accurate 3D Model Generation System

National Technical Research Organisation (NTRO) · Robotics and Drones · Software

Genuinely hard and near the research frontier, with the real data arriving only at the event — attempt it only if you know photogrammetry, and prove metric accuracy against a measured reference rather than showing a good-looking mesh, because that is exactly where single-pass geometry falls short.

Data: Will be provided real time.

What it actually is

Building an accurate 3D model from drone footage normally needs multiple overlapping passes and heavy post-processing, but in disaster response or reconnaissance there is often only one flight over the target. The ask is a system that reconstructs a georeferenced, metrically accurate 3D model from a single drone pass video — terrain, buildings, roads and vegetation — despite the limited viewing angles that single pass provides.

What to build

A reconstruction system taking a single-pass drone video with GPS and flight metadata and producing a georeferenced, metrically accurate textured 3D model — point cloud or mesh — of terrain, building facades and rooftops, roads and vegetation, handling the hard constraints the description lists: limited viewing angles from one path, motion blur and compression artefacts, variable illumination, dynamic objects, GPS noise, occluded surfaces and near-real-time processing, using structure-from-motion and multi-view stereo augmented with learned depth or neural reconstruction to compensate for the sparse geometry a single pass provides.

Smallest thing that wins the room

Process a single-pass drone video clip end to end into a textured 3D model, then measure a building dimension on the model and show it matching the real dimension within a stated tolerance, proving the reconstruction is metric rather than merely visual.

How crowded this one gets

A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.

Moderate110–250 teams expectedroughly 1 in 91–209 wins it

Quieter than 64% of the 226 · #81 of 226 by expected field

A normal-sized field. Your idea has to be good, not miraculous.

Why: defence, intelligence and space bodies drew small fields.

This is a guess, not a fact

Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.

The scores

The number is the shorthand. The line under it is the reason.

What you will be writing

  • COLMAP structure-from-motion + multi-view stereo
  • Neural reconstruction (NeRF / 3D Gaussian Splatting)
  • Learned monocular depth priors
  • Georeferencing from GPS/IMU flight metadata
  • Open3D meshing and texturing
  • Metric accuracy evaluation against reference
  • Photogrammetry
  • 3D reconstruction
  • Drone mapping

Prior art to read before you start

single-pass 3D reconstruction · metric drone photogrammetry · neural scene reconstruction

Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.