AI-Based Detection and Classification of Industrial Fires and Persistent Thermal Sources Using NASA FIRMS, OSM & Satellite Data

The receipts 🧾

Every red flag on this statement, in full. These are the four places it bites.

1. Exhibit A

   FIRMS resolution is coarse at 375 metres to 1 kilometre, so precisely attributing a hotspot to a specific facility is uncertain in dense industrial areas

2. It gets worse

   Ground-truth labels for anomaly classes are scarce, so validating your classifier honestly is difficult and you may lean on land-cover as a proxy for truth

3. Still reading?

   Agricultural burning is seasonal and location-overlapping with other classes, so seasonal context is needed or it is easily confused

4. And the finisher

   Persistent industrial thermal sources and accidental fires can co-locate, so distinguishing a flare-up at a refinery from its normal flare is genuinely hard

The damage report 📉

Every score this statement earned, and what each one actually costs you.

• Feasibility

  4/5

  Actually buildable, which on this slate is rarer than it sounds. Do not squander it on scope.

  NASA FIRMS is free and linked, OpenStreetMap industrial facility data and land-cover products are open, and the classification is a tractable problem of combining location context with the thermal detection's temporal persistence — all achievable with standard geospatial tooling.

• Innovation scope

  3/5

  Mildly interesting. The novelty will not carry the room; the build has to.

  The classification approach is fairly determined by the available context layers — location plus land cover plus persistence largely decides the class — so your room is in handling ambiguous cases and in the persistence analysis rather than in the concept.

• Clarity

  4/5

  The ask is unambiguous, which quietly removes your favourite excuse.

  The description names the data sources, the categories to distinguish, and the specific deliverable of separating industrial from natural fires with a GIS overlay, so the requirement is well defined.

• Acceptance potential

  4/5

  Strong footing before you have written a line. Try not to waste it.

  Underrated — all data is free and linked, the build is genuinely tractable, the classification is a real gap in FIRMS-based monitoring, and the geospatial-plus-thermal domain thins the field while the NTRO framing gives it clear purpose.

• Effort

  Heavy

  Heavy. Somebody on this team is not sleeping in week three. Pick who, on purpose.

  FIRMS ingestion, contextual layer integration, the classification logic and temporal persistence analysis plus a GIS front end are focused, well-bounded pieces on open data.

• Demo-ability

  Easy

  Easy to demo — and so is everyone else's. Working is the floor here, not the achievement.

  Classified hotspots on a map, each justified by its facility or land-cover context, is immediately legible and the refinery-flare-versus-crop-burning contrast tells the story instantly.

The demo they will have already seen 🪞

Somewhere around 150–340 teams are heading here, and the description is doing the choosing for most of them. They will read the same brief, reach the same architecture, and build a version of the same demo you are planning. Being correct is the floor. If your five minutes could be swapped with the team before you and nobody in the room would notice, you have not picked badly — you have built predictably, which costs exactly the same and hurts more.

What survives 🛡️

The ground worth standing on when the questions start.

• NASA FIRMS is free and linked directly, and the supporting OSM and land-cover layers are all open, so the entire pipeline runs on free data
• Distinguishing a persistent gas flare from a transient accidental fire by temporal behaviour is a genuinely clever, buildable feature
• Classified hotspots justified by facility and land-cover context make an immediately convincing map demo