Research map / Detection and segmentation / Weakly supervised detection and segmentation
Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 top-down attention cues (ablation) | Weakly supervised segmentation model · PASCAL VOC 2012 validation and test | mIoU (%): 30.5 -> 52.8; 31.6 -> 53.7 | Helped |
| Add the attention-based segmentation loss (ablation) | Weakly supervised segmentation model · PASCAL VOC 2012 validation and test | mIoU (%): 47.5 -> 52.8; 49.1 -> 53.7 | Helped |
| Add attention cues to CRF unary potentials (ablation) | CRF-RNN segmentation refinement · PASCAL VOC 2012 validation and test | mIoU (%): 50.1 -> 52.8; 51.4 -> 53.7 | Helped |
| Add neighboring-neuron dependencies to attention estimation (ablation) | Top-down attention estimation · PASCAL VOC 2012 validation and test | mIoU (%): 51.3 -> 52.8; 52.1 -> 53.7 | Helped |
| Use CRF-RNN instead of dense CRF post-processing (ablation) | Segmentation refinement · PASCAL VOC 2012 validation and test | mIoU (%): 49.4 -> 52.8; 51.3 -> 53.7 | Helped |
| Add attention cues to inference-time CRF unary potentials (ablation) | CRF-RNN inference · PASCAL VOC 2012 validation and test | mIoU (%): 50.4 -> 52.8; 51.8 -> 53.7 | Helped |
| Use a learned fully connected layer rather than global max pooling (ablation) | Image-level score aggregation · PASCAL VOC 2012 validation and test | mIoU (%): 48.3 -> 52.8; 49.6 -> 53.7 | Helped |
| Use a learned fully connected layer rather than global average pooling (ablation) | Image-level score aggregation · PASCAL VOC 2012 validation and test | mIoU (%): 47.4 -> 52.8; 48.3 -> 53.7 | Helped |
| Use a learned fully connected layer rather than log-sum-exponential pooling (ablation) | Image-level score aggregation · PASCAL VOC 2012 validation and test | mIoU (%): 50.8 -> 52.8; 51.3 -> 53.7 | Helped |
| Fuse bottom-up segmentation, top-down attention, and CRF refinement | Weakly supervised segmentation · PASCAL VOC 2012 validation and test | mIoU (%): 50.7 -> 52.8; 51.7 -> 53.7 | Helped |
Closest papers by meaning
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.93)
- Webly Supervised Semantic Segmentation (CVPR 2017 · similarity 0.92)
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.92)
- Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning (ICLR 2021 · similarity 0.92)
- Learning Pixel-Level Semantic Affinity With Image-Level Supervision for Weakly Supervised Semantic Segmentation (CVPR 2018 · similarity 0.91)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.90)
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.