Research map / Detection and segmentation / CNN features
Active Learning Meets Foundation Models: Fast Remote Sensing Data Annotation for Object Detection
Marvin Burges, Philipe Ambrozio Dias, Carson Woody, Sarah Walters, Dalton Lunga
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 |
|---|---|---|---|
| add dual-source uncertainty estimation instead of entropy sampling (ablation) | AL4FM-OD sampler · DIOR (100 labeled images) | mAP: 0.4±0.4 -> 3.8±0.5 | Improved |
| use entropy-based sampling instead of random sampling (ablation) | active learning sampler · DIOR (100 labeled images) | mAP: 1.6±0.6 -> 0.4±0.4 | Hurt |
| add mask-guided diversity estimation after DSUE (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 5.4±0.8 -> 10.7±0.7 | Improved |
| replace CoreSet diversity stage with MDE (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 4.8±0.6 -> 10.7±0.7 | Improved |
| replace DivProto diversity stage with MDE (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 0.9±0.4 -> 10.7±0.7 | Improved |
| replace PPAL CCMS diversity stage with MDE (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 5.2±0.8 -> 10.7±0.7 | Improved |
| increase budget expansion ratio from 1 to 6 (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 5.4±0.8 -> 10.7±0.7 | Improved |
| increase expansion ratio from 6 to 7 (ablation) | AL4FM-OD · DIOR | selection time: 45s -> 87s | Hurt |
| use frozen DinoV2 backbone instead of FMOW ResNet50 (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 10.7±0.7 -> 15.4±0.8 | Improved |
| use ScaleMAE backbone instead of ResNet50 (ablation) | AL4FM-OD · DIOR (300 labeled images) | mAP: 10.7±0.7 -> 4.3±0.9 | Hurt |
| add Dynamic Box Switching (SAM boxes replace predictions) (ablation) | RTDETRv2 detector · Waffle Homes (first annotation round) | Recall@100 at IoU 0.75: 4× improvement | Improved |
| add Dynamic Box Switching (ablation) | RTDETRv2 detector · HRSC2016 (first annotation round) | Recall@100 at IoU 0.75: 2× improvement | Improved |
Closest papers by meaning
- Dual Domain Control via Active Learning for Remote Sensing Domain Incremental Object Detection (ICCV 2025 · similarity 0.91)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.89)
- Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community (AAAI 2025 · similarity 0.88)
- ROS-SAM: High-Quality Interactive Segmentation for Remote Sensing Moving Object (CVPR 2025 · similarity 0.87)
- Semi-supervised Open-World Object Detection (AAAI 2024 · similarity 0.87)
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.86)
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.
