Research map / Detection and segmentation / CNN features
Linear Context Transform Block
Read the paperCheck 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 |
|---|---|---|---|
| add LCT block | ResNet50 · ImageNet-1K val | Top-1 accuracy (%): 76.15 -> 77.45 | Helped |
| replace SE with LCT | ResNet50 · ImageNet-1K val | Top-1 accuracy (%): 77.31 -> 77.45 | Helped |
| add LCT block | ResNet101 · ImageNet-1K val | Top-1 accuracy (%): 77.37 -> 78.55 | Helped |
| add LCT block | Faster R-CNN with ResNet101 · COCO 2017 val detection | AP0.5:0.95 bbox: 38.5 -> 40.0 | Helped |
| add LCT block | Mask R-CNN with ResNet101 · COCO 2017 val detection | AP0.5:0.95 bbox: 39.4 -> 40.9 | Helped |
| add LCT block | Cascade R-CNN with ResNet101 · COCO 2017 val detection | AP0.5:0.95 bbox: 42.0 -> 43.6 | Helped |
| add LCT block | Cascade Mask R-CNN with ResNet101 · COCO 2017 val detection | AP0.5:0.95 bbox: 42.6 -> 44.1 | Helped |
| add LCT block | Mask R-CNN with ResNet101 · COCO 2017 val instance segmentation | AP0.5:0.95 mask: 35.9 -> 37.0 | Helped |
| add LCT block | Cascade Mask R-CNN with ResNet101 · COCO 2017 val instance segmentation | AP0.5:0.95 mask: 37.0 -> 38.1 | Helped |
| use different group counts G=1, 4, 8, 16, 32, or 64 (ablation) | LCT-ResNet50 · ImageNet-1K val | Top-1 accuracy (%): 77.37; 77.36; 77.44; 77.34; 77.32; 77.45 | No clear effect |
| remove group normalization (ablation) | LCT-ResNet50 · ImageNet-1K val | Top-1 accuracy (%): 77.45 -> 76.89 | Hurt |
| remove linear transform (ablation) | LCT-ResNet50 · ImageNet-1K val | Top-1 accuracy (%): 77.45 -> 76.82 | Hurt |
Closest papers by meaning
- XAttention: Block Sparse Attention with Antidiagonal Scoring (ICML 2025 · similarity 0.86)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · similarity 0.85)
- Attention (as Discrete-Time Markov) Chains (NeurIPS 2025 · similarity 0.84)
- xT: Nested Tokenization for Larger Context in Large Images (ICML 2024 · similarity 0.84)
- CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up (NeurIPS 2025 · similarity 0.84)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · 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.