Research map / Vision-language and multimodal / Multimodal LLMs
LogicSeg: Parsing Visual Semantics with Neural Logic Learning and Reasoning
Liulei Li, Wenguan Wang, Yi Yang
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 logic-induced training with C, D and E rule losses (ablation) | DeepLabV3+ ResNet-101 · Mapillary Vistas 2.0 val (level 1) | mIoU: 37.43 -> 40.72 | Improved |
| add composition rule loss L_C alone (ablation) | DeepLabV3+ ResNet-101 · Mapillary Vistas 2.0 val (level 1) | mIoU: 37.43 -> 38.71 | Improved |
| add decomposition rule loss L_D alone (ablation) | DeepLabV3+ ResNet-101 · Mapillary Vistas 2.0 val (level 1) | mIoU: 37.43 -> 38.98 | Improved |
| add exclusion rule loss L_E alone (ablation) | DeepLabV3+ ResNet-101 · Mapillary Vistas 2.0 val (level 1) | mIoU: 37.43 -> 38.43 | Improved |
| add logic-induced inference with 2 message passing iterations (ablation) | LogicSeg DeepLabV3+ · Mapillary Vistas 2.0 val (level 1) | mIoU: 40.18 -> 40.72 | Improved |
| increase message passing iterations from 2 to 4 (ablation) | LogicSeg inference · Mapillary Vistas 2.0 val (level 1) | mIoU: 40.72 -> 40.03 | Hurt |
| vary aggregation coefficient q from 5 to 10 or 1 (ablation) | forall quantifier approximation · Mapillary Vistas 2.0 val (level 1) | mIoU: 40.72 -> 38.25 (q=10), 38.77 (q=1) | Hurt |
| apply LogicSeg to DeepLabV3+ and MaskFormer across datasets | segmentation models · Mapillary, Cityscapes, Pascal-Part-108, ADE20K | mIoU: +1.12% to +3.29% mIoU | Improved |
| add logic-induced training | DeepLabV3+ · Mapillary Vistas 2.0 training speed | training time (min/epoch): 45.62 -> 47.67 (about 5%) | Hurt |
| add logic-induced inference | DeepLabV3+ · Mapillary Vistas 2.0 inference speed | fps: 3.44 -> 3.31 | Hurt |
| LogicSeg on Cityscapes with Swin-S MaskFormer vs Hssn and SegFormer | MaskFormer · Cityscapes val (level 1) | mIoU: 83.85 vs SegFormer 83.90, Hssn 83.74 | No clear effect |
Closest papers by meaning
- Logic-induced Diagnostic Reasoning for Semi-supervised Semantic Segmentation (ICCV 2023 · similarity 0.90)
- CoReS: Orchestrating the Dance of Reasoning and Segmentation (ECCV 2024 · similarity 0.88)
- HyperSeg: Hybrid Segmentation Assistant with Fine-grained Visual Perceiver (CVPR 2025 · similarity 0.87)
- Seg-ReSearch: Segmentation with Interleaved Reasoning and External Search (ICML 2026 · similarity 0.87)
- Integrating Low-Level Visual Cues for Enhanced Unsupervised Semantic Segmentation (AAAI 2025 · similarity 0.86)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.86)
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.
