Research map / Detection and segmentation / Weakly supervised detection and segmentation
Enabling Deep Residual Networks for Weakly Supervised Object Detection
Yunhang Shen, Rongrong Ji, Yan Wang, Zhiwei Chen, Feng Zheng, Feiyue Huang, Yunsheng Wu
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 redundant adaptation neck (RAN) (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 37.5 | Improved |
| add small-kernel convolution stem (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 32.9 | Improved |
| add MaxPool down-sampling (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 33.7 | Improved |
| add dilated convolution (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 30.9 | Improved |
| combine RAN, small-kernel, MaxPool and dilated convolution (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 43.4 | Improved |
| combine all four principles (ablation) | OICR on ResNet18 · PASCAL VOC 2007 test | mAP: 34.7 -> 51.0 | Improved |
| combine all four principles (ablation) | PCL on ResNet18 · PASCAL VOC 2007 test | mAP: NAN -> 50.2 | Improved |
| combine all four principles (ablation) | C-MIL on ResNet18 · PASCAL VOC 2007 test | mAP: NAN -> 52.6 | Improved |
| combine all four principles (ablation) | WSDDN on ResNet101 · PASCAL VOC 2007 test | mAP: 25.7 -> 44.1 | Improved |
| freeze four backbone stages instead of none (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 25.2 -> 27.5 | Improved |
| freeze all five backbone stages instead of four (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 27.5 -> 14.4 | Hurt |
| use RoIAlign instead of RoIPool (ablation) | WSDDN on ResNet18 · PASCAL VOC 2007 test | mAP: 28.4 -> 24.2 | Hurt |
Closest papers by meaning
- W2N: Switching from Weak Supervision to Noisy Supervision for Object Detection (ECCV 2022 · similarity 0.91)
- UWSOD: Toward Fully-Supervised-Level Capacity Weakly Supervised Object Detection (NeurIPS 2020 · similarity 0.91)
- WSOD2: Learning Bottom-Up and Top-Down Objectness Distillation for Weakly-Supervised Object Detection (ICCV 2019 · similarity 0.90)
- PCL: Proposal Cluster Learning for Weakly Supervised Object Detection (TPAMI 2018 · similarity 0.90)
- Learning an Invariant and Equivariant Network for Weakly Supervised Object Detection (TPAMI 2023 · similarity 0.89)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 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.
