Research map / Detection and segmentation / CNN features
Active Learning for Deep Object Detection via Probabilistic Modeling
Jiwoong Choi, Ismail Elezi, Hyuk-Jae Lee, Clement Farabet, Jose M. Alvarez
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 |
|---|---|---|---|
| replace SSD with a mixture density model on localization and classification heads (ablation) | SSD · VOC07 | mAP IoU>0.75: 43.36 -> 46.11 | Improved |
| replace SSD with a mixture density model on localization and classification heads (ablation) | SSD · MS-COCO | mAP IoU>0.75: 11.93 -> 13.57 | Improved |
| use max of four uncertainties instead of random sampling (ablation) | active learning on SSD · VOC07 | mAP at 4k images: 68.47 -> 69.43 | Improved |
| combine localization and classification uncertainties rather than one task alone (ablation) | active learning on SSD · VOC07 | image selection overlap: 48% -> 14% | Improved |
| use max instead of sum of the four uncertainties (ablation) | active learning on SSD · VOC07 | mAP at 4k images: 69.09 -> 69.43 | Improved |
| use Ours_gmm sampling instead of MC-dropout | active learning · VOC07 | mAP at 4k images: 69.39 -> 69.43 | Improved |
| add Ours_eff scoring instead of LLAL on MS-COCO | active learning · MS-COCO | mAP at 7k images: 29.53 -> 30.02 | Improved |
| increase mixture components K from 1 to 4 (ablation) | Ours_gmm · VOC07 | mAP IoU>0.75: 45.18 -> 46.11 | Improved |
| increase mixture components K from 4 to 8 (ablation) | Ours_gmm · VOC07 | mAP IoU>0.75: 46.11 -> 45.69 | Hurt |
| raise input resolution from 300 to 512 (ablation) | Ours_gmm · VOC07 | mAP IoU>0.75: 46.11 -> 48.23 | Improved |
| use Type-2 instead of Type-1 classification loss (ablation) | Ours_gmm · MS-COCO | mAP at 7k images: 30.51 -> 29.55 | Hurt |
| use Type-2 instead of Type-1 classification loss (ablation) | Ours_gmm · VOC07 | mAP at 4k images: 69.43 -> 69.40 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Faster-RCNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Not All Labels Are Equal: Rationalizing the Labeling Costs for Training Object Detection (CVPR 2022 · similarity 0.91)
- Active Learning for Object Detection with Evidential Deep Learning and Hierarchical Uncertainty Aggregation (ICLR 2023 · similarity 0.90)
- Portable Active Learning for Object Detection (CVPR 2026 · similarity 0.90)
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.90)
- Active Domain Adaptation with False Negative Prediction for Object Detection (CVPR 2024 · similarity 0.87)
- Uncertainty Meets Diversity: A Comprehensive Active Learning Framework for Indoor 3D Object Detection (CVPR 2025 · 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.
