Research map / Detection and segmentation / Open-set and few-shot detection
ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic 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 |
|---|---|---|---|
| Replace direct pixel-wise entropy minimization with adversarial alignment of weighted self-information | Deeplab-V2 segmentation models · GTA5→Cityscapes | mIoU (%): 32.8 -> 36.1; 42.3 -> 43.8 | Helped |
| Add entropy minimization to the top 30% highest-entropy target pixels (ablation) | ResNet-101 MinEnt · GTA5→Cityscapes | mIoU (%): 42.3 -> 43.1 | Helped |
| Add adversarial entropy adaptation to direct entropy minimization (ablation) | ResNet-101 GTA5→Cityscapes model · GTA5→Cityscapes | mIoU (%): 43.8 -> 45.5 | Helped |
| Add a source class-ratio prior to direct entropy minimization (ablation) | VGG-16 MinEnt · SYNTHIA→Cityscapes (16 classes) | mIoU (%): 27.5 -> 30.4; +2.9% | Helped |
| Add a source class-ratio prior to direct entropy minimization (ablation) | VGG-16 MinEnt · SYNTHIA→Cityscapes (13 classes) | mIoU* (%): 32.5 -> 35.4; +2.9% | Helped |
| Add adversarial entropy adaptation to the VGG-16 class-prior MinEnt setting (ablation) | VGG-16 SYNTHIA→Cityscapes model · SYNTHIA→Cityscapes (16 classes) | mIoU (%): 30.4 -> 31.4; ~+1% | Helped |
| Add adversarial entropy adaptation to the VGG-16 class-prior MinEnt setting (ablation) | VGG-16 SYNTHIA→Cityscapes model · SYNTHIA→Cityscapes (13 classes) | mIoU* (%): 35.4 -> 36.6; ~+1% | Helped |
| Add a source class-ratio prior to MinEnt (ablation) | VGG-16 SYNTHIA→Cityscapes model · SYNTHIA→Cityscapes, road class | IoU: 37.8 -> 45.9 | Helped |
| Combine direct entropy minimization and adversarial entropy adaptation (ablation) | ResNet-101 model · SYNTHIA→Cityscapes (16 classes) | mIoU (%): 40.8 -> 41.2 | Helped |
| Combine direct entropy minimization and adversarial entropy adaptation (ablation) | ResNet-101 model · SYNTHIA→Cityscapes (13 classes) | mIoU* (%): 47.6 -> 48.0 | Helped |
| Apply entropy-based adaptation to the SSD-300 detection model | SSD-300 · Cityscapes→Cityscapes Foggy | mAP (%): 14.7 -> 16.9 | Helped |
| Use adversarial entropy adaptation rather than direct entropy minimization | SSD-300 · Cityscapes→Cityscapes Foggy | mAP (%): 14.7 -> 26.2; +11.5% | Helped |
Closest papers by meaning
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.90)
- DAST: Unsupervised Domain Adaptation in Semantic Segmentation Based on Discriminator Attention and Self-Training (AAAI 2021 · similarity 0.90)
- Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation (CVPR 2018 · similarity 0.89)
- Joint Adversarial Learning for Domain Adaptation in Semantic Segmentation (AAAI 2020 · similarity 0.89)
- Learning to Adapt Structured Output Space for Semantic Segmentation (CVPR 2018 · similarity 0.89)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.87)
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.