Research map / Detection and segmentation / Weakly supervised detection and segmentation
Background Activation Suppression for Weakly Supervised Object Localization and Semantic Segmentation
Wei Zhai, Pingyu Wu, Kai Zhu, Yang Hua Cao, Feng Wu, Zheng-Jun Zha
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 background activation suppression loss L_bas (ablation) | FPM baseline · CUB-200-2011 | GT-known Loc: 71.14 -> 92.15 | Improved |
| add L_bas (ablation) | FPM baseline · CUB-200-2011 | Top-1 Loc: 57.89 -> 74.15 | Improved |
| add Top-k strategy for final map (ablation) | BAS · CUB-200-2011 | GT-known Loc: 92.15 -> 95.41 | Improved |
| add L_bas (ablation) | WSSS baseline · PASCAL VOC 2012 | mIoU: 50.1 -> 57.7 | Improved |
| add L_bas (ablation) | WSSS baseline · MS COCO 2014 | mIoU: 32.5 -> 36.9 | Improved |
| insert generator after layer 3 instead of layer 1 (ablation) | BAS · CUB-200-2011 | GT-known Loc: 52.87 -> 95.41 | Improved |
| insert generator after layer 3 instead of layer 4 (ablation) | BAS · CUB-200-2011 | GT-known Loc: 90.94 -> 95.41 | Improved |
| insert generator after layer 3 instead of layer 4 (ablation) | BAS · PASCAL VOC 2012 | mIoU: 55.3 -> 57.7 | Improved |
| insert generator after layer 3 instead of layer 4 (ablation) | BAS · MS COCO 2014 | mIoU: 34.8 -> 36.9 | Improved |
| remove final ReLU and apply ReLU separately before S and S^b (journal vs conference version) | BAS · CUB-200-2011 mean over 4 backbones | GT-known Loc: +0.83 | Improved |
| same ReLU change | BAS · ILSVRC mean over 4 backbones | GT-known Loc: +0.11 | Improved |
| mask feature maps instead of original image for background map (ablation) | BAS · CUB-200-2011 | localization accuracy | Improved |
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.
- C2AM — C2AM: Contrastive Learning of Class-Agnostic Activation Map for Weakly Supervised Object Localization and Semantic Segmentation (CVPR 2022) · better on 1 of 1 shared result
- CLIMS — CLIMS: Cross Language Image Matching for Weakly Supervised Semantic Segmentation (CVPR 2022) · better on 1 of 1 shared result
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
- Background Activation Suppression for Weakly Supervised Object Localization (CVPR 2022 · similarity 0.95)
- Online Refinement of Low-Level Feature Based Activation Map for Weakly Supervised Object Localization (ICCV 2021 · similarity 0.90)
- Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration (ECCV 2022 · similarity 0.89)
- CREAM: Weakly Supervised Object Localization via Class RE-Activation Mapping (CVPR 2022 · similarity 0.89)
- Background-Aware Classification Activation Map for Weakly Supervised Object Localization (TPAMI 2023 · similarity 0.89)
- Built-in Foreground/Background Prior for Weakly-Supervised Semantic Segmentation (ECCV 2016 · similarity 0.89)
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.
