Research map / Architectures and efficiency / CNN design
Factorized Bilinear Models for Image Recognition
Yanghao Li, Naiyan Wang, Jiaying Liu, Xiaodi Hou
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 |
|---|---|---|---|
| add a factorized bilinear layer before global pooling | Inception-BN · CIFAR-100 | test error (%): 24.70 -> 21.98 | Improved |
| add FB layer in the lower In5a factory (ablation) | Inception-BN · CIFAR-100 | test error (%): 24.70 -> 24.73 | No clear effect |
| apply FB per spatial position (Conv-FBN) instead of in the final FC layer (ablation) | Inception-BN · CIFAR-100 | test error (%): 24.07 -> 21.98 | Improved |
| add a second FB block (In5b) on top of Conv-FBN (ablation) | Inception-BN · CIFAR-100 | test error (%): 21.98 -> 23.70 | Hurt |
| increase number of factors k from 10 to 20 (ablation) | Conv-FBN · CIFAR-100 | test error (%): 22.99 -> 21.98 | Improved |
| increase number of factors k from 20 to 80 (ablation) | Conv-FBN · CIFAR-100 | test error (%): 21.98 -> 21.88 | No clear effect |
| add DropFactor (ablation) | FBN (Inception-BN) · CIFAR-100 | test error (%): 23.55 -> 21.98 | Improved |
| add Dropout on top of DropFactor (ablation) | FBN (Inception-BN) · CIFAR-100 | test error (%): 21.98 -> 22.71 | Hurt |
| use 3x3 instead of 1x1 FB kernel (ablation) | Conv-FBN · CIFAR-100 | test error (%): 21.98 -> 23.08 | Hurt |
| add FB layer (Conv-FBN) | ResNet-1001 · CIFAR-100 | top-1 error (%): 20.50 -> 19.67 | Improved |
| add FB layer (Conv-FBN) | ResNet-1001 · CIFAR-10 | top-1 error (%): 4.04 -> 4.09 | No clear effect |
| add FB layer (Conv-FBN) | ResNet-50 · ImageNet val | top-1 error (%): 24.7 -> 24.0 | Improved |
Closest papers by meaning
- Bilinear Convolutional Neural Networks for Fine-Grained Visual Recognition (TPAMI 2017 · similarity 0.87)
- Learning Deep Bilinear Transformation for Fine-grained Image Representation (NeurIPS 2019 · similarity 0.86)
- Low-Rank Bilinear Pooling for Fine-Grained Classification (CVPR 2017 · similarity 0.86)
- Revisiting Bilinear Pooling: A Coding Perspective (AAAI 2020 · similarity 0.85)
- Multilinear Operator Networks (ICLR 2024 · similarity 0.85)
- Star with Bilinear Mapping (CVPR 2025 · similarity 0.82)
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.
