Research map / Detection and segmentation / CNN features
On-Orbit Real-Time Wildfire Detection Under On-Board Constraints
Matthias Rötzer, Veronika Pörtge, Martin Ickerott, Jayendra Chorapalli, Dimitri Scheftelowitsch, Max Bereczky, Dmitry Rashkovetsky, Sai Appalla, Julia Gottfriedsen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| initialize encoder with DenseMAE pretraining instead of random init (ablation) | DenseMAE emb64 + lightweight head · OTC-P1 MWIR test | AP: 0.551 -> 0.689 | Improved |
| fine-tune encoder end-to-end instead of freezing it (ablation) | DenseMAE emb64 · OTC-P1 MWIR test | AP: 0.612 -> 0.689 | Improved |
| DenseMAE pretraining vs supervised U-Net++ baseline | DenseMAE emb64 + head · OTC-P1 MWIR test | AP: 0.650 -> 0.689 | Improved |
| increase embedding dimension from 32 to 64 (ablation) | DenseMAE downstream models · OTC-P1 MWIR test | AP: 0.671 -> 0.689 AP; latency 65.34 -> 110.46 ms (TRT-Head) | Mixed |
| add high-resolution refinement head (ablation) | DenseMAE DW-Res-Head emb64 · OTC-P1 MWIR test | Fire-F1: 0.711 -> 0.744 | Improved |
| add high-resolution refinement head (ablation) | DenseMAE DW-Res-Head emb64 · OTC-P1 MWIR test | AP: 0.689 -> 0.699 | Improved |
| use TRT-Head instead of U-Net++ baseline (emb32) | DenseMAE emb32 · OTC-P1 MWIR test | latency: 104.34 -> 65.34 ms | Improved |
| add EMA distillation to MAE pretraining (ablation) | DenseMAE · OTC-P1 MWIR test (SSL ablation) | Test AP: 0.1612 -> 0.2796 (D=64) in Table 2, but text says gains marginal and inconsistent | Mixed |
| raise EMA distillation weight from 0.02 to 0.5 (ablation) | Hybrid DenseMAE D=32 b=4 · OTC-P1 MWIR test (SSL ablation) | Test AP: 0.255 -> 0.1828 | Hurt |
| reduce mask block size from 8 to 2 (ablation) | DenseMAE D=64 r=0.60 · OTC-P1 MWIR test (SSL ablation) | Test AP: 0.1251 -> 0.1812 | Improved |
| increase embedding dimension from 32 to 128 in MAE pretraining (ablation) | DenseMAE r=0.60 b=4 · OTC-P1 MWIR test (SSL ablation) | Test AP: 0.1137 -> 0.1593 | Improved |
Closest papers by meaning
- SmokeViz: A Large-Scale Satellite Dataset for Wildfire Smoke Detection and Segmentation (NeurIPS 2025 · similarity 0.84)
- DetectiumFire: A Comprehensive Multi-modal Dataset Bridging Vision and Language for Fire Understanding (NeurIPS 2025 · similarity 0.82)
- Resource-Aware Parameter-Efficient Model Adaptation for Onboard High-Dimensional Data (NeurIPS 2026 · similarity 0.82)
- FireScope: Wildfire Risk Raster Prediction With a Chain-of-Thought Oracle (CVPR 2026 · similarity 0.81)
- Hotspotter: A Generalizable Pipeline for Automated Detection of Subtle Volcanic Thermal Features in Satellite Images (AAAI 2025 · similarity 0.81)
- Federated Oriented Learning: A Practical One-Shot Personalized Federated Learning Framework (ICML 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
