Research map / Architectures and efficiency / CNN design
Grassmann Pooling as Compact Homogeneous Bilinear Pooling for Fine-Grained Visual Classification
Xing Wei, Yue Zhang, Yihong Gong, Jiawei Zhang, Nanning Zheng
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 full bilinear pooling with Grassmann pooling GP-256 | FBP · FGVC Aircraft | accuracy (%): 86.9 -> 89.8 | Improved |
| replace full bilinear pooling with Grassmann pooling GP-256 | FBP · Food-101 | accuracy (%): 82.4 -> 85.7 | Improved |
| replace improved bilinear pooling (matrix square root) with GP-256 | IBP · Stanford Cars | accuracy (%): 92.0 -> 92.8 | Improved |
| replace improved bilinear pooling with GP-256 | IBP · CUB Bird-200 | accuracy (%): 85.8 -> 85.8 | No clear effect |
| replace improved bilinear pooling with GP-256 | IBP · Food-101 | accuracy (%): 84.2 -> 85.7 | Improved |
| replace low-rank bilinear pooling LRBP with GP-64 | LRBP · FGVC Aircraft | accuracy (%): 87.3 -> 88.1 | Improved |
| replace low-rank bilinear pooling LRBP with GP-64 | LRBP · CUB Bird-200 | accuracy (%): 84.2 -> 85.4 | Improved |
| replace compact bilinear CBP-TS with GP-64 | CBP-TS · Stanford Cars | accuracy (%): 91.2 -> 91.7 | Mixed |
| reduce channels from 256 to 64 (GP-256 to GP-64) (ablation) | Grassmann pooling · FGVC Aircraft | accuracy (%): 89.8 -> 88.1 | Hurt |
| replace 1x1 conv reduction by halving feature maps (512 to 256) (ablation) | Grassmann pooling k=16 · FGVC Aircraft | accuracy (%): 89.8 -> 89.2 | Hurt |
| add bilinear pooling to the VGG-16 FC baseline | VGG-16 FC · FGVC Aircraft | accuracy (%): 74.1 -> 86.9 | Improved |
Closest papers by meaning
- Learning Compact Discriminant Representation via Low-Rank Bilinear Pooling (TPAMI 2025 · similarity 0.91)
- Compact Bilinear Pooling (CVPR 2016 · similarity 0.91)
- Low-Rank Bilinear Pooling for Fine-Grained Classification (CVPR 2017 · similarity 0.90)
- Efficient Compact Bilinear Pooling via Kronecker Product (AAAI 2022 · similarity 0.90)
- Hierarchical Bilinear Pooling for Fine-Grained Visual Recognition (ECCV 2018 · similarity 0.89)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.85)
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.
