Research map / Detection and segmentation / CNN features
Chained Cascade Network for Object Detection
Wanli Ouyang, Kun Wang, Xin Zhu, Xiaogang Wang
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 2 extra cascade stages with classifier chaining (ablation) | BN-Net+FRCN · ImageNet val2 | mAP: 49.4 -> 50.5 | Improved |
| add 4 extra cascade stages with classifier chaining (ablation) | BN-Net+FRCN · ImageNet val2 | mAP: 49.4 -> 50.9 | Improved |
| add classifier chaining (use previous stage scores) to 4 extra cascade stages (ablation) | BN-Net cascade · ImageNet val2 | mAP: 50.5 -> 50.9 | Improved |
| add cascade chaining on top of multi-region features (ablation) | BN-Net · ImageNet val2 | mAP: 50.5 -> 51.3 | Improved |
| add cascade chaining on top of multi-region features (ablation) | BN-Net · VOC07 test | mAP: 73.8 -> 75.2 | Improved |
| add feature chaining on top of cascade and classifier chaining (ablation) | BN-Net · ImageNet val2 | mAP: 51.3 -> 54.5 | Improved |
| add feature chaining on top of cascade and classifier chaining (ablation) | BN-Net · VOC07 test | mAP: 75.2 -> 77.1 | Improved |
| add multi-region features without chaining or cascade (ablation) | BN-Net · VOC07 test | mAP: 70.8 -> 73.8 | Improved |
| add multi-region features without chaining or cascade (ablation) | BN-Net · ImageNet val2 | mAP: 49.4 -> 50.5 | Improved |
| learn scaling vectors a and b instead of fixing them to 1 (ablation) | CC-Net · ImageNet val2 | mAP: drop by 1.7% | Improved |
| replace scaling vector a with fully connected layer with C1C2 parameters (ablation) | CC-Net feature chaining · ImageNet val2 | mAP | No clear effect |
| CC-Net feature chaining vs GBD-Net message passing when all features share the same contextual region | BN-Net · ImageNet val2 | mAP: 50.2% -> 52% | Improved |
Closest papers by meaning
- Weakly Supervised Cascaded Convolutional Networks (CVPR 2017 · similarity 0.87)
- CRAFT Objects From Images (CVPR 2016 · similarity 0.85)
- Cascade RPN: Delving into High-Quality Region Proposal Network with Adaptive Convolution (NeurIPS 2019 · similarity 0.85)
- Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018 · similarity 0.84)
- Bi-Directional Cascade Network for Perceptual Edge Detection (CVPR 2019 · similarity 0.84)
- BDCN: Bi-Directional Cascade Network for Perceptual Edge Detection (TPAMI 2020 · 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.
