Research map / Architectures and efficiency / RNNs and normalization
BNET: Batch Normalization With Enhanced Linear Transformation
Yuhui Xu, Lingxi Xie, Cihang Xie, Wenrui Dai, Jieru Mei, Siyuan Qiao, Wei Shen, Hongkai Xiong, Alan Yuille
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 |
|---|---|---|---|
| replace BN with BNET-3 | ResNet-50 · ImageNet | top-1 accuracy: 76.3 -> 76.8 | Improved |
| replace BN with BNET-3 | MobileNetV2 0.25 · ImageNet | top-1 accuracy: 53.6 -> 55.9 | Improved |
| replace BN with BNET-3 | Faster R-CNN ResNet-50 · MS-COCO | AP: 37.5 -> 39.5 | Improved |
| enlarge BNET kernel from 3 to 7 | Faster R-CNN ResNet-50 · MS-COCO | AP: 39.5 -> 40.7 | Improved |
| replace BN with BNET-3 | TSN ResNet-50 · UCF-101 | top-1 accuracy: 79.8 -> 84.1 | Improved |
| enlarge BNET kernel from 3 to 5 | TSN ResNet-50 · UCF-101 | top-1 accuracy: 84.1 -> 84.2 | No clear effect |
| replace BN with BNET-3 under AutoAugment 100 epochs | ResNet-50 · ImageNet | top-1 accuracy: 76.1 -> 77.2 | Improved |
| use 5x5 instead of 3x3 kernel in BNET (ablation) | ResNet-50 · ImageNet | top-1 accuracy: 76.8 -> 76.6 | Hurt |
| use dilation 2 for 3x3 BNET kernel (ablation) | ResNet-50 · ImageNet | top-1 accuracy: 76.8 -> 76.5 | Hurt |
| apply BNET at position C rather than A (ablation) | ResNet-50 · ImageNet | top-1 accuracy: 76.5 -> 76.8 | Improved |
| apply BNET to all three BN layers instead of only C (ablation) | ResNet-50 · ImageNet | top-1 accuracy: 76.8 -> 76.7 | Hurt |
| use BNET instead of adding a depth-wise conv layer (ablation) | ResNet-50 · ImageNet | top-1 accuracy: 75.7 -> 76.8 | Improved |
Closest papers by meaning
- Understanding Batch Normalization (NeurIPS 2018 · similarity 0.93)
- Revisiting Batch Norm Initialization (ECCV 2022 · similarity 0.91)
- Re-Thinking the Effectiveness of Batch Normalization and Beyond (TPAMI 2023 · similarity 0.91)
- Normalization Propagation: A Parametric Technique for Removing Internal Covariate Shift in Deep Networks (ICML 2016 · similarity 0.90)
- Reducing Divergence in Batch Normalization for Domain Adaptation (AAAI 2025 · similarity 0.88)
- Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion (ICLR 2024 · 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.
