Research map / Architectures and efficiency / CNN design
Cross-Convolutional-Layer Pooling for Image Recognition
Lingqiao Liu, Chunhua Shen, Anton van den Hengel
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 cross-layer pooling instead of SCFV on OConv, AlexNet | SCFV · MIT-67 | accuracy: 59.2 -> 63.0 | Improved |
| use cross-layer pooling instead of SCFV on OConv, VGGNet | SCFV · Birds-200 | accuracy: 73.7 -> 77.0 | Improved |
| use AConv layer instead of OConv for cross-layer pooling on Birds-200 (ablation) | CrossLayer VGGNet · Birds-200 | accuracy: 77.0 -> 69.4 | Hurt |
| use AConv layer instead of OConv for cross-layer pooling on PASCAL-07 (ablation) | CrossLayer VGGNet · PASCAL VOC 2007 | mAP: 84.1 -> 87.1 | Improved |
| add PCA to the full normalization pipeline (ablation) | CrossLayer AConv MIT-67 · MIT-67 | accuracy: 72.6 -> 78.2 | Improved |
| add power normalization (ablation) | CrossLayer AConv MIT-67 · MIT-67 | accuracy: 76.4 -> 78.2 | Improved |
| add l2 normalization (ablation) | CrossLayer AConv MIT-67 · MIT-67 | accuracy: 77.1 -> 78.2 | Improved |
| use deeper layers conv5-3/conv5-4 instead of conv5-2/conv5-3 (ablation) | CrossLayer VGGNet · PASCAL VOC 2007 | mAP: 82.2 -> 84.1 | Improved |
| use sum-pooling instead of max-pooling for channel pooling (ablation) | CrossLayer VGGNet · MIT-67 | accuracy: 72.8 -> 74.4 | Improved |
| use max-pooling instead of sum-pooling for channel pooling (ablation) | CrossLayer VGGNet · PASCAL VOC 2007 | mAP: 84.1 -> 85.2 | Mixed |
| apply feature sign quantization (2 bits per dimension) (ablation) | CrossLayer VGGNet · PASCAL VOC 2007 | mAP: 87.0 -> 85.1 | Hurt |
| apply feature sign quantization (2 bits per dimension) (ablation) | CrossLayer VGGNet · MIT-67 | accuracy: 78.2 -> 77.9 | No clear effect |
Closest papers by meaning
- Fine-Tuning CNN Image Retrieval with No Human Annotation (TPAMI 2018 · similarity 0.89)
- Wavelet Pooling for Convolutional Neural Networks (ICLR 2018 · similarity 0.88)
- Kernel Pooling for Convolutional Neural Networks (CVPR 2017 · similarity 0.88)
- End-to-End Learning of Deep Visual Representations for Image Retrieval (IJCV 2017 · similarity 0.88)
- Convolutional Patch Representations for Image Retrieval: An Unsupervised Approach (IJCV 2016 · similarity 0.88)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 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.
