Research map / Detection and segmentation / Weakly supervised detection and segmentation
Information Entropy Based Feature Pooling for Convolutional Neural Networks
Weitao Wan, Jiansheng Chen, Tianpeng Li, Yiqing Huang, Jingqi Tian, Cheng Yu, Youze Xue
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 |
|---|---|---|---|
| use EP at inference only on pre-trained ResNet-50 (ablation) | ResNet-50 · ImageNet val (10-crop) | top-1 error (%): 22.85 -> 22.73 | Improved |
| train with EP instead of inference-only EP (ResNet-50) (ablation) | ResNet-50 with EP inference · ImageNet val (10-crop) | top-1 error (%): 22.73 -> 22.64 | Improved |
| add multi-level BEW layers to EP (ResNet-50) (ablation) | ResNet-50 + EP · ImageNet val (10-crop) | top-1 error (%): 22.64 -> 22.45 | Improved |
| add multi-level BEW to VGG-16 | VGG-16 · ImageNet val (10-crop) | top-1 error (%): 28.07 -> 24.78 | Improved |
| replace GAP with EP and add BEW in VGG-GAP | VGG-GAP · ImageNet val (10-crop) | top-1 error (%): 31.66 -> 30.04 | Improved |
| add multi-level BEW + EP to InceptionV3 | InceptionV3 · ImageNet val (10-crop) | top-1 error (%): 20.20 -> 19.92 | Improved |
| add EP to iSQRT-COV pooling (ResNet-50) | iSQRT-COV · ImageNet val (one-crop) | top-1 error (%): 22.14 -> 21.97 | Improved |
| add BEW + EP to SENet (ResNet-152) | SENet · ImageNet val (one-crop) | top-1 error (%): 21.57 -> 21.08 | Improved |
| add BEW to VGG-16 for scene recognition | VGG-16 · Places365 val (10-crop) | top-1 error (%): 44.76 -> 43.36 | Improved |
| use ResNet-50 + BEW + EP versus deeper ResNet-152 | ResNet-152 · Places365 val (10-crop) | top-1 error (%): 45.26 -> 43.75 | Improved |
| use entropy weights instead of CAM for localization (ResNet-50) | ResNet-50 + CAM · ImageNet val localization | top-1 localization error (%): 51.12 -> 50.15 | Improved |
| add EP to DSRG seed generation | DSRG · VOC 2012 val | mIoU (%): 59.0 -> 61.5 | Improved |
Closest papers by meaning
- Reducing Information Bottleneck for Weakly Supervised Semantic Segmentation (NeurIPS 2021 · similarity 0.87)
- Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation (CVPR 2021 · similarity 0.86)
- Learning to Predict Context-adaptive Convolution for Semantic Segmentation (ECCV 2020 · similarity 0.86)
- Building Detail-Sensitive Semantic Segmentation Networks With Polynomial Pooling (CVPR 2019 · similarity 0.86)
- Loss Max-Pooling for Semantic Image Segmentation (CVPR 2017 · similarity 0.86)
- Global Pooling, More Than Meets the Eye: Position Information Is Encoded Channel-Wise in CNNs (ICCV 2021 · 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.
