Research map / Detection and segmentation / Weakly supervised detection and segmentation
FickleNet: Weakly and Semi-Supervised Semantic Image Segmentation Using Stochastic Inference
Jungbeom Lee, Eunji Kim, Sungmin Lee, Jangho Lee, Sungroh Yoon
Check 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 |
|---|---|---|---|
| Use FickleNet localization maps instead of DSRG maps | Weakly supervised segmentation with DeepLab-VGG16 · PASCAL VOC 2012 val | mIoU: 59.0 -> 61.2 | Improved |
| Use FickleNet localization maps instead of DSRG maps | Weakly supervised segmentation with DeepLab-VGG16 · PASCAL VOC 2012 test | mIoU: 60.4 -> 61.9 | Improved |
| Use FickleNet localization maps instead of DSRG maps | Weakly supervised segmentation with DeepLab-v2-ResNet101 · PASCAL VOC 2012 val | mIoU: 61.4 -> 64.9 | Improved |
| Use FickleNet localization maps instead of DSRG maps | Weakly supervised segmentation with DeepLab-v2-ResNet101 · PASCAL VOC 2012 test | mIoU: 63.2 -> 65.3 | Improved |
| Use FickleNet with 1.4K strong and 9K weak training data instead of DSRG | Semi-supervised segmentation · PASCAL VOC 2012 val | mIoU: 64.3 -> 65.8 | Improved |
| Add map expansion to the naive implementation (ablation) | FickleNet training · Training iterations | time per iteration: 20 sec/iter -> 1.3 sec/iter | Improved |
| Add map expansion to the naive implementation (ablation) | FickleNet localization-map extraction · CAM extraction | time per image: 2.98 sec/img -> 0.21 sec/img | Improved |
| Use stochastic selection in both training and inference instead of general dropout in both (ablation) | FickleNet · PASCAL VOC 2012 val | mIoU: 49.1 -> 61.2 | Improved |
| Use stochastic selection during inference rather than deterministic inference after stochastic training (ablation) | FickleNet · PASCAL VOC 2012 val | mIoU: 59.6 -> 61.2 | Improved |
| Use stochastic selection during training rather than deterministic training with deterministic inference (ablation) | FickleNet · PASCAL VOC 2012 val | mIoU: 59.0 -> 61.2 | Improved |
| Increase FickleNet dropout rate from 0.3 to 0.9 (ablation) | FickleNet · PASCAL VOC 2012 val | mIoU: 58.8 -> 61.2 | Improved |
| Replace location-wise hidden-unit selection with general dropout at rate 0.9 (ablation) | FickleNet localization and segmentation · PASCAL VOC 2012 val | mIoU: 61.2 -> 49.1 | Hurt |
Later papers that compared with this method
10 later accepted papers report FickleNet as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- Label-Efficient Hybrid-Supervised Learning for Medical Image Segmentation (AAAI 2022 · better on 1 of 1 shared result)
- ECS-Net: Improving Weakly Supervised Semantic Segmentation by Using Connections Between Class Activation Maps (ICCV 2021 · better on 1 of 1 shared result)
- Employing Multi-Estimations for Weakly-Supervised Semantic Segmentation (ECCV 2020 · better on 4 of 4 shared results)
- Self-Supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation (CVPR 2020 · better on 1 of 2 shared results)
- Weakly-Supervised Semantic Segmentation via Sub-Category Exploration (CVPR 2020 · better on 2 of 2 shared results)
- Learning Saliency-Free Model with Generic Features for Weakly-Supervised Semantic Segmentation (AAAI 2020 · better on 2 of 2 shared results)
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · better on 2 of 2 shared results)
- Splitting vs. Merging: Mining Object Regions with Discrepancy and Intersection Loss for Weakly Supervised Semantic Segmentation (ECCV 2020 · better on 2 of 2 shared results)
2 more not listed.
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
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.89)
- Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation (ECCV 2016 · similarity 0.89)
- Modeling the Background for Incremental and Weakly-Supervised Semantic Segmentation (TPAMI 2021 · similarity 0.88)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.88)
- Semi-supervised Semantic Segmentation via Strong-weak Dual-branch Network (ECCV 2020 · similarity 0.88)
- Weakly Supervised Semantic Segmentation with Boundary Exploration (ECCV 2020 · 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.
