Research map / Detection and segmentation / Weakly supervised detection and segmentation
Integral Object Mining via Online Attention Accumulation
Peng-Tao Jiang, Qibin Hou, Yang Cao, Ming-Ming Cheng, Yunchao Wei, Hong-Kai Xiong
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 OAA accumulation with average fusion to CAM (ablation) | CAM · VOC12 val | mIoU: 53.9% -> 57.0% | Improved |
| add OAA accumulation with maximum fusion to CAM (ablation) | CAM · VOC12 val | mIoU: 53.9% -> 58.6% | Improved |
| use maximum rather than average fusion in OAA (ablation) | OAA · VOC12 val | mIoU: 57.0% -> 58.6% | Improved |
| use hybrid loss instead of multi-label cross-entropy (ablation) | OAA+ · VOC12 val | mIoU: 51.2% -> 59.6% | Improved |
| train integral attention model (OAA+ with hybrid loss) on top of OAA (ablation) | OAA · VOC12 val | mIoU: 58.6% -> 59.6% | Improved |
| use multi-label cross-entropy for the integral attention model instead of using OAA alone (ablation) | OAA · VOC12 val | mIoU: 58.6% -> 51.2% | Hurt |
| increase segmentation training images from 20% to 100% (ablation) | OAA+ · VOC12 val | mIoU: 54.6% -> 59.6% | Improved |
| use only 20% of training images with OAA+ labels versus CAM with all images (ablation) | OAA+ · VOC12 val | mIoU: 53.9% -> 54.6% | Improved |
| switch backbone from VGG16 to ResNet-101 | OAA+ · VOC12 val | mIoU: 63.1% -> 65.2% | Improved |
| use OAA+ attention instead of GAIN erasing | segmentation pipeline · VOC12 val | mIoU: 55.3% -> 63.1% | Improved |
Closest papers by meaning
- Online Attention Accumulation for Weakly Supervised Semantic Segmentation (TPAMI 2021 · similarity 0.96)
- L2G: A Simple Local-to-Global Knowledge Transfer Framework for Weakly Supervised Semantic Segmentation (CVPR 2022 · similarity 0.90)
- Mining Cross-Image Semantics for Weakly Supervised Semantic Segmentation (ECCV 2020 · similarity 0.86)
- Looking Beyond Single Images for Weakly Supervised Semantic Segmentation Learning (TPAMI 2022 · similarity 0.86)
- Expectation-Maximization Attention Networks for Semantic Segmentation (ICCV 2019 · similarity 0.86)
- Training-Free Open-Ended Object Detection and Segmentation via Attention as Prompts (NeurIPS 2024 · 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.
