Research map / Detection and segmentation / CNN features
Local Context Normalization: Revisiting Local Normalization
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 BN with local context normalization | HRNetV2 W48 · Cityscapes val | mIoU Class (%): 76.22 -> 77.49 | Helped |
| Replace BN with local context normalization | HRNetV2 W18 Small v1 · Cityscapes val | mIoU Class (%): 71.27 -> 71.77 | Helped |
| Use LCN instead of GN | Mask R-CNN R50 · COCO val2017 (minival) | AP^bbox (%): 37.34 -> 37.90 | Helped |
| Use LCN instead of GN | Mask R-CNN R50 · COCO val2017 (minival) | AP^mask (%): 34.33 -> 34.50 | Helped |
| Replace BN with LCN | Resnet 50 · ImageNet val | Top 1 Err. (%): 23.59 -> 24.23 | Hurt |
| Replace BN with LCN | U-Net · INRIA Aerial Image Labeling test | Overall IoU: 67.98 -> 69.90 | Helped |
| Replace BN with LCN | UNet · Maryland 2013 test tiles | mIoU (%): 76.69 -> 76.51 | Hurt |
| Increase the LCN window toward the full spatial extent (ablation) | HRNetV2 Small v1 · Cityscapes val | mIoU Class (%): 71.77 -> 70.03 | Hurt |
| Set LCN channels per group to 4 instead of 2 (ablation) | HRNetV2 W18 Small v1 · Cityscapes val | mIoU Class (%): 71.77 -> 70.26 | Hurt |
| Set LCN channels per group to 8 instead of 2 (ablation) | HRNetV2 W18 Small v1 · Cityscapes val | mIoU Class (%): 71.77 -> 70.14 | Hurt |
| Set LCN channels per group to 16 instead of 2 (ablation) | HRNetV2 W18 Small v1 · Cityscapes val | mIoU Class (%): 71.77 -> 70.11 | Hurt |
Closest papers by meaning
- Dynamic Token Normalization improves Vision Transformers (ICLR 2022 · similarity 0.87)
- Kalman Normalization: Normalizing Internal Representations Across Network Layers (NeurIPS 2018 · similarity 0.87)
- Normalize Filters! Classical Wisdom for Deep Vision (NeurIPS 2025 · similarity 0.85)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.85)
- Role of Locality and Weight Sharing in Image-Based Tasks: A Sample Complexity Separation between CNNs, LCNs, and FCNs (ICLR 2024 · similarity 0.84)
- Dynamic Group Normalization: Spatio-Temporal Adaptation to Evolving Data Statistics (CVPR 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.