AI Hackathon
← DEPARTURES
PK1124AI Video & ImageBoarding

BTAJI CREW · OBSCURA: Satellite AI Research Challenge

BTAJI CREW · on Devpost

Cash prize: $50
Free credits: Credits: not stated
Who can enter: Solo OK Country rules: unknown
Code rights: Not checked yet
Submissions closeNov 24

Student teams use satellite images and environmental data to estimate air pollution and fire signals around industrial sites.

Proceed to gate Opens obscura.devpost.com. You register on the organizer's site.

Imported from Devpost and approved by our team. We haven't checked the rules by hand. Confirm the rules on the event page before you enter.

One email a week

New events that match your topics, every Monday. Unsubscribe anytime.

Flight details

Submissions close:
2026-11-24 15:00 UTC
Event ends:
2026-11-24 15:00 UTC
Cash prize:
$50

Prizes listed: Total cash prizes: USD 50 (3 cash, 1 other)

In short

  • Submit by November 24, 2026, but confirm the dates on the organizer page because they come from a third-party listing.
  • The total cash prize pool is USD 50, and digital participation certificates are also listed.
  • Models must generalize to unseen industrial sites and time periods; the page lists students only, so check the rules.

Project ideas

Cloud-Aware Methane Estimator

A model that flags methane hotspots around industrial sites while learning to ignore cloud gaps and missing labels.

Architecture: Load Sentinel-2 chips and 29 weather and terrain features. → Mask cloudy pixels and impute missing satellite readings. → Train a multi-target regressor with masked RMSE loss. → Validate on held-out facilities and report per-target error.

Stack: Python, PyTorch, Sentinel-2, scikit-learn

Unseen-Site NO2 Generalizer

A site-aware validation pipeline that tests whether nitrogen dioxide predictions transfer to facilities outside the training set.

Architecture: Group samples by facility and observation window. → Build image and tabular feature encoders. → Train with grouped cross-validation splits. → Compare per-site errors and flag weak regions.

Stack: Python, PyTorch, pandas, scikit-learn

Fire Signal Fusion Model

A multimodal model that fuses thermal fire signals with smoke, aerosol and satellite imagery to estimate flare activity.

Architecture: Join Sentinel-2 chips with thermal and aerosol targets. → Encode imagery and tabular features separately. → Fuse representations and predict fire activity. → Audit performance on sparse and missing labels.

Stack: Python, PyTorch, rasterio, pandas

FAQ

Can I enter alone?
The event page says individuals or teams of up to 20 students may enter, so solo entry is allowed. Confirm team rules on the official page.
Who is eligible to participate?
The event page lists students above the legal age of majority, excludes companies, and opens all countries except standard exceptions. Eligibility is not confirmed, so check the official rules before entering.

Sources

Open the event page →

More open AI hackathons