Research map / Detection and segmentation / CNN features
Revisiting the Sibling Head in Object Detector
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the shared proposal with task-specific spatial proposals (ablation) | Faster R-CNN sibling head · COCO minival | AP: 36.1 -> 38.2 | Helped |
| Add progressive constraint to TSD (ablation) | TSD · COCO minival | AP: 38.2 -> 41.0 | Helped |
| Add joint training with the sibling head (ablation) | TSD · COCO minival | AP: 38.2 -> 39.7 | Helped |
| Disentangle tasks in the backbone rather than spatially at the detection head (ablation) | ResNet-50 Faster R-CNN · COCO minival | AP: 36.1 -> 22.3; 22.0; 20.3 | Hurt |
| Use separate task-specific heads without task-specific spatial proposals (ablation) | ResNet-50 Faster R-CNN · COCO minival | AP: 36.1 -> 37.3 | Helped |
| Add a classification margin constraint (ablation) | TSD jointly trained with sibling head · COCO minival | AP: 39.7 -> 40.1 | Helped |
| Add a localization margin constraint (ablation) | TSD jointly trained with sibling head · COCO minival | AP: 39.7 -> 40.8 | Helped |
| Use pointwise deformation for localization instead of proposal-wise translation (ablation) | TSD with pointwise classification proposals, without progressive constraint · COCO minival | AP: 38.2 -> 38.5 | Helped |
| Use pointwise deformation for classification and proposal-wise translation for localization (ablation) | TSD with progressive constraint · COCO minival | AP: 39.8 -> 41.0 | Helped |
| Add TSD and progressive constraint | Faster R-CNN · COCO minival | AP: 36.1 -> 41.0 | Helped |
| Add TSD | Faster R-CNN · OpenImage validation | AP.5: 64.64 -> 68.18 | Helped |
| Add TSD and progressive constraint | Faster R-CNN with ResNet-101-DCN · COCO minival | AP: 40.8 -> 43.5 | Helped |
Closest papers by meaning
- Rethinking Classification and Localization for Object Detection (CVPR 2020 · similarity 0.88)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.85)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (CVPR 2024 · similarity 0.84)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.84)
- SparseAlign: a Fully Sparse Framework for Cooperative Object Detection (CVPR 2025 · similarity 0.84)
- Union-over-Intersections: Object Detection beyond Winner-Takes-All (ICLR 2025 · similarity 0.84)
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.