Research map / Label-efficient and robust learning / Semi-supervised learning
Semi Supervised Semantic Segmentation Using Generative Adversarial Network
Nasim Souly, Concetto Spampinato, Mubarak Shah
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 unlabeled data via GAN semi-supervised training | FCN-VGG16 baseline · VOC 2012 val (all labeled + unlabeled) | mean IU: 59.5 -> 64.1 | Improved |
| add unlabeled data via GAN semi-supervised training | FCN-VGG16 baseline · VOC 2012 val (all labeled + unlabeled) | mean accuracy: 69.2 -> 80.7 | Improved |
| add unlabeled data via GAN semi-supervised training | FCN-VGG16 baseline · VOC 2012 val (all labeled + unlabeled) | pixel accuracy: 89.9 -> 90.5 | No clear effect |
| add image-level class conditioning (weak supervision) to semi-supervised GAN | Semi Supervised GAN · VOC 2012 val (all labeled + unlabeled) | mean IU: 64.1 -> 65.8 | Improved |
| add image-level class conditioning (weak supervision) to semi-supervised GAN | Semi Supervised GAN · VOC 2012 val (all labeled + unlabeled) | mean accuracy: 80.7 -> 80.0 | No clear effect |
| add unlabeled data via GAN semi-supervised training | Fully supervised · VOC 2012 val (30% labeled) | mean IU: 38.9 -> 42.2 | Improved |
| add weak supervision | Fully supervised · VOC 2012 val (30% labeled) | mean IU: 38.9 -> 44.6 | Improved |
| add unlabeled data via GAN semi-supervised training | Fully supervised · SiftFlow (2000 unlabeled SUN2012 images) | mean IU: 34.4 -> 35.1 | Improved |
| add unlabeled data with 20% labeled data | Full · VOC 2012 val (20% labeled) | mean IU: 16.0 -> 19.8 | Improved |
| add unlabeled data with 50% labeled data | Full · VOC 2012 val (50% labeled) | pixel accuracy: 88.5 -> 88.4 | No clear effect |
| add unlabeled data via GAN semi-supervised training | Fully supervised · StanfordBG (10k unlabeled PASCAL images) | mean IU: 53.1 -> 63.3 | Improved |
| add unlabeled frames via GAN semi-supervised training | Ours Fully supervised · CamVid (11k unlabeled frames) | pixel accuracy: 88.4 -> 87.0 | Hurt |
Closest papers by meaning
- Semantic Segmentation With Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization (CVPR 2021 · similarity 0.93)
- Repurposing GANs for One-Shot Semantic Part Segmentation (CVPR 2021 · similarity 0.90)
- Weakly Supervised Semantic Segmentation via Adversarial Learning of Classifier and Reconstructor (CVPR 2023 · similarity 0.90)
- DatasetGAN: Efficient Labeled Data Factory With Minimal Human Effort (CVPR 2021 · similarity 0.90)
- Transferable Semi-Supervised Semantic Segmentation (AAAI 2018 · similarity 0.90)
- KE-GAN: Knowledge Embedded Generative Adversarial Networks for Semi-Supervised Scene Parsing (CVPR 2019 · 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.
