Research map / Detection and segmentation / CNN features
Disentangle then Parse: Night-time Semantic Segmentation with Illumination Disentanglement
Zhixiang Wei, Lin Chen, Tao Tu, Pengyang Ling, Huaian Chen, Yi Jin
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 DTP (disentangle then parse) to UPer-Swin | UPer-Swin · NightCity-fine val | mIoU: 61.1 -> 64.2 | Improved |
| add DTP to HRNetV2, DeepLabV3+, ConvNeXt, Swin segmenters | day-time segmentation methods · NightCity-fine val | mIoU: +1.5 to +3.1 | Improved |
| add disentangle, retinex and smooth losses together (ablation) | SOD · NightCity-fine val | mIoU: 61.1 -> 62.5 | Improved |
| add semantic disentanglement loss L_SeDe (ablation) | SOD · NightCity-fine val | mIoU: 62.5 -> 63.7 | Improved |
| add illumination segmentation loss L_segill in IAParser (ablation) | DTP · NightCity-fine val | mIoU: 63.7 -> 64.2 | Improved |
| add random illumination noise disturb operation (ablation) | SOD · NightCity-fine val | mIoU: 62.5 -> 64.2 | Improved |
| use simple preprocessing (histogram normalization or CLAHE with grayscale) instead of learned disentanglement (ablation) | baseline UPer-Swin · NightCity-fine val | mIoU: 61.1 -> 62.0 | No clear effect |
| use a larger disentanglement model (Base to Huge) (ablation) | DTP · NightCity-fine val | mIoU: 64.2 -> 64.2 | No clear effect |
| train on refined NightCity-fine labels instead of original NightCity (ablation) | Swin · NightCity val | mIoU: 58.4 -> 60.0 | Improved |
| add DTP to Swin on BDD100K-night with small training set | UPer-Swin · BDD100K-night | mIoU: 31.7 -> 36.9 | Improved |
Closest papers by meaning
- Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation (ICCV 2019 · similarity 0.91)
- The Parables of the Mustard Seed and the Yeast: Extremely Low-Budget, High-Performance Nighttime Semantic Segmentation (AAAI 2025 · similarity 0.89)
- NightAdapter: Learning a Frequency Adapter for Generalizable Night-time Scene Segmentation (CVPR 2025 · similarity 0.87)
- See through the Dark: Learning Illumination-affined Representations for Nighttime Occupancy Prediction (NeurIPS 2025 · similarity 0.86)
- Exploring the Common Appearance-Boundary Adaptation for Nighttime Optical Flow (ICLR 2024 · similarity 0.85)
- Integrating Low-Level Visual Cues for Enhanced Unsupervised Semantic Segmentation (AAAI 2025 · similarity 0.84)
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.
