Research map / Detection and segmentation / Weakly supervised detection and segmentation
LabelEnc: A New Intermediate Supervision Method for Object Detection
Miao Hao, Yitao Liu, Xiangyu Zhang, Jian Sun
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 LabelEnc intermediate supervision (vs 1x baseline) | RetinaNet ResNet50 · COCO minival | mmAP: 36.4 -> 38.6 | Improved |
| add LabelEnc intermediate supervision (vs 1x baseline) | FCOS ResNet50 · COCO minival | mmAP: 38.7 -> 41.1 | Improved |
| add LabelEnc intermediate supervision (vs 1x baseline) | FPN ResNet50 · COCO minival | mmAP: 36.8 -> 38.6 | Improved |
| add LabelEnc compared with doubled-schedule 2x baseline | RetinaNet Res101-DCN · COCO minival | mmAP: 41.5 -> 42.9 | Improved |
| train Step 1 with joint detection and distance losses instead of reconstruction loss only (ablation) | LabelEnc RetinaNet ResNet50 · COCO minival | mmAP: 37.6 -> 38.6 | Improved |
| train Step 1 with joint losses instead of reconstruction loss only (ablation) | LabelEnc RetinaNet ResNet101 · COCO minival | mmAP: 39.7 -> 40.5 | Improved |
| use ResNet101 instead of ResNet50 as Step 1 auxiliary backbone (ablation) | LabelEnc RetinaNet ResNet101 · COCO minival | mmAP: 40.5 -> 40.6 | No clear effect |
| add Step 2 on top of Step 1 alone (ablation) | LabelEnc RetinaNet ResNet50 · COCO minival | mmAP: 37.5 -> 38.6 | Improved |
| add Step 2 on top of Step 1 alone (ablation) | LabelEnc FCOS ResNet50 · COCO minival | mmAP: 40.3 -> 41.1 | Improved |
| add detection head initialization from Step 1 (with supervision) (ablation) | LabelEnc RetinaNet ResNet50 · COCO minival | mmAP: 37.2 -> 38.6 | Improved |
| add intermediate supervision (with head initialization) (ablation) | LabelEnc RetinaNet ResNet50 · COCO minival | mmAP: 37.7 -> 38.6 | Improved |
| replace ResNet50 teacher KD with label encoding function | RetinaNet ResNet50 · COCO minival | mmAP: 37.3 -> 38.6 | Improved |
Closest papers by meaning
- Diverse Learner: Exploring Diverse Supervision for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.89)
- Not All Labels Are Equal: Rationalizing the Labeling Costs for Training Object Detection (CVPR 2022 · similarity 0.89)
- LEOD: Label-Efficient Object Detection for Event Cameras (CVPR 2024 · similarity 0.88)
- Learning to Overcome Noise in Weak Caption Supervision for Object Detection (TPAMI 2022 · similarity 0.88)
- Many-shot from Low-shot: Learning to Annotate using Mixed Supervision for Object Detection (ECCV 2020 · similarity 0.88)
- Co-Student: Collaborating Strong and Weak Students for Sparsely Annotated Object Detection (ECCV 2024 · similarity 0.88)
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.
