An interactive web application that accepts satellite images and natural-language queries, using an agentic AI backend to route tasks to specialist models for single-image, cross-modal, and change-detection analysis.
Indian Space Research Organisation(ISRO) · Department of Space / Indian Space Research Organisation · Software
Signing in saves it for your whole team, so everyone on your invite link sees the same two entries. Anything you shortlisted while signed out comes with you.
Decomposed from what the description asks for. Nothing added.
Input Upload and Validation
An interface module that accepts GeoTIFF, TIFF, or benchmark images and checks their format, metadata, and compatibility.
Remote-Sensing Adapted VLM
A vision-language component fine-tuned on datasets like BigEarthNet to handle remote-sensing imagery and terminology.
Specialist Tool Registry
A collection of task-specific models handling visual question answering, captioning, grounding, change analysis, and optical-SAR fusion.
Agentic Controller
An orchestration layer that interprets natural-language queries, selects appropriate tools, sequences workflows, and compiles execution summaries.
Interactive GUI
A web application interface displaying textual results, visual evidence, confidence information, and downloadable reports.
Both columns are read off the brief's own wording. Nothing here is inferred from the ministry's name.
The jury will evaluate your system using prescribed public benchmark test subsets and a hidden ISRO/SAC evaluation dataset containing Cartosat-2S optical and RISAT SAR image pairs. They will check if your system successfully performs single-image VQA, an additional single-image task, multitemporal change analysis, optical-SAR paired analysis, and provides an auditable execution summary.
A jury can still ask about these. Decide them deliberately rather than by accident.
Generated from the brief's own wording and the competition's published rules, never from a guess about what this ministry prefers.
Specific accuracy or performance metric thresholds required to pass the evaluation?
The brief never answers this, so a panel will. Whatever you decide, say it the same way twice.
Hardware specifications of the deployment environment for the final demo?
The brief never answers this, so a panel will. Whatever you decide, say it the same way twice.
Has any part of this been shown at a previous event, hackathon or college project?
The guidelines are explicit: your solution must not have appeared in any previous event or programme, of any sort. A recycled project is what a team under time pressure reaches for.
The brief asks for input upload and validation. How would you build that?
Decoded from SIH26167 itself: An interface module that accepts GeoTIFF, TIFF, or benchmark images and checks their format, metadata, and compatibility. The brief asks for it by name.
The brief asks for remote-sensing adapted vlm. How would you build that?
Decoded from SIH26167 itself: A vision-language component fine-tuned on datasets like BigEarthNet to handle remote-sensing imagery and terminology. The brief asks for it by name.
The brief asks for specialist tool registry. How would you build that?
Decoded from SIH26167 itself: A collection of task-specific models handling visual question answering, captioning, grounding, change analysis, and optical-SAR fusion. The brief asks for it by name.
The brief asks for agentic controller. How would you build that?
Decoded from SIH26167 itself: An orchestration layer that interprets natural-language queries, selects appropriate tools, sequences workflows, and compiles execution summaries. The brief asks for it by name.
The brief asks for interactive gui. How would you build that?
Decoded from SIH26167 itself: A web application interface displaying textual results, visual evidence, confidence information, and downloadable reports. The brief asks for it by name.
Why not use what already exists? Name the closest thing to this that is already running.
A team that has not named the alternative themselves is answering this for the first time in the room.
Which single thing will you demonstrate end to end, start to finish, with nothing skipped?
Ours, not a rule: a narrow thing that fully works survives questioning better than a broad thing that half works. If nobody on the team can name it, that is the finding.
Show me this working: the jury will evaluate your system using prescribed public benchmark test subsets and a hidden ISRO/SAC evaluation dataset containing Cartosat-2S optical and RISAT SAR image pairs.
This is the evaluator read for your problem statement, decoded from the brief's own wording.
Show me this working: they will check if your system successfully performs single-image VQA, an additional single-image task, multitemporal change analysis, optical-SAR paired analysis, and provides an auditable execution summary.
This is the evaluator read for your problem statement, decoded from the brief's own wording.
Each one is quoted from a gap in the brief, not a guess about your team.
Needs data you may not get
Final evaluation will use a hidden ISRO/SAC evaluation dataset containing pre-georeferenced and co-registered Cartosat-2S optical and RISAT SAR image pairs that are not disclosed to participating teams.
What the organisers attached, and what the brief assumes you can get.
Same organisation, same year. Reading two of theirs tells you more about what they care about than reading one.