Research map / Detection and segmentation / Open-set and few-shot detection
Cornerformer: Purifying Instances for Corner-Based Detectors
Haoran Wei, Xin Chen, Lingxi Xie, Qi Tian
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 corner pooling with CTE (ablation) | CornerNet · COCO val2017 (CornerNet) | AP (%): 39.1 -> 40.8 | Improved |
| replace Soft-NMS with A3-NMS (ablation) | CornerNet with CTE · COCO val2017 (CornerNet) | AP (%): 40.8 -> 41.3 | Improved |
| replace Soft-NMS with A3-NMS (ablation) | CornerNet with CTE · Citypersons val | AP50 (%): 40.3 -> 45.4 | Improved |
| add full Cornerformer (CTE + A3-NMS) (ablation) | CornerNet · COCO val2017 | AP (%): 36.9 -> 41.3 | Improved |
| add Cornerformer | CornerNet · COCO test-dev | AP (%): 40.5 -> 42.6 | Improved |
| add Cornerformer | CenterNet · COCO test-dev | AP (%): 44.9 -> 46.1 | Improved |
| add Cornerformer with pull-push branch | CentripetalNet · COCO test-dev | AP (%): 46.1 -> 47.1 | Improved |
| use two separate CTEs for the two corner types | CornerNet with Cornerformer · COCO test-dev | AP (%): 42.6 -> 43.5 | Improved |
| replace corner pooling with CTE (faster speed) | CornerNet · COCO test-dev | FPS: 4.1 -> 7.4 | Improved |
| add positional encoding to criss-cross attention (PCCA) (ablation) | CTE · COCO val2017 (CornerNet) | AP (%): 39.3 -> 40.1 | Improved |
| add output positional encoding PE (ablation) | CTE with PCCA · COCO val2017 (CornerNet) | AP (%): 40.1 -> 40.8 | Improved |
| use 4 CTE blocks instead of 3 (ablation) | CornerNet w/ CTE · COCO val2017 (CornerNet) | AP (%): 40.8 -> 40.4 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CentripetalNet — CentripetalNet: Pursuing High-Quality Keypoint Pairs for Object Detection (CVPR 2020) · better on 2 of 2 shared results
- CornerNet — CornerNet: Detecting Objects as Paired Keypoints (ECCV 2018) · better on 2 of 2 shared results
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
- CornerNet: Detecting Objects as Paired Keypoints (ECCV 2018 · similarity 0.88)
- Corner Proposal Network for Anchor-free, Two-stage Object Detection (ECCV 2020 · similarity 0.87)
- AShapeFormer: Semantics-Guided Object-Level Active Shape Encoding for 3D Object Detection via Transformers (CVPR 2023 · similarity 0.86)
- CentripetalNet: Pursuing High-Quality Keypoint Pairs for Object Detection (CVPR 2020 · similarity 0.86)
- HumanLiker: A Human-like Object Detector to Model the Manual Labeling Process (NeurIPS 2022 · similarity 0.85)
- CenterNet: Keypoint Triplets for Object Detection (ICCV 2019 · similarity 0.85)
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.
