Research map / Label-efficient and robust learning / Recognition and long-tail
Compact Bilinear Pooling
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Tensor Sketch compact bilinear pooling instead of full bilinear pooling | VGG-M · CUB-200-2011, no fine-tuning | classification error (%): 29.41 -> 31.53 | Hurt |
| Use Tensor Sketch compact bilinear pooling instead of full bilinear pooling | VGG-D · CUB-200-2011, fine-tuned | classification error (%): 16.00 -> 16.00 | No clear effect |
| Use Tensor Sketch compact bilinear pooling instead of full bilinear pooling | VGG-M · MIT Indoor, no fine-tuning | classification error (%): 29.77 -> 30.71 | Hurt |
| Fine-tune the Tensor Sketch model | VGG-M · MIT Indoor | classification error (%): 30.71 -> 31.30 | Helped |
| Fine-tune the full and compact bilinear models | VGG-D · MIT Indoor | classification error (%): 22.45 -> 28.98; 23.83 -> 27.27 | Hurt |
| Use Tensor Sketch compact bilinear pooling instead of full bilinear pooling | VGG-D · DTD | classification error (%): 32.50 -> 32.29; 35.04 -> 35.49 | Mixed |
| Use Tensor Sketch rather than full bilinear features | VGG-M · CUB-200-2011, 1 image/class | mean average precision (%): 12.7 -> 15.6 | Helped |
| Reduce Tensor Sketch dimension from 16K to 512 (ablation) | Tensor Sketch, CUB-200 · CUB-200, fine-tuned | classification error (%): 22.5 -> 25.54 | Hurt |
| Fine-tune the compact pooling model (ablation) | Tensor Sketch, CUB-200, d=32 · CUB-200 | classification error (%) | Helped |
| Tune the projection parameters W (ablation) | Compact bilinear pooling · CUB-200 | classification error (%) | Helped |
| Use Tensor Sketch rather than PCA-reduced bilinear features | CUB-200, fine-tuned · CUB-200, 16K dimensions | Top-1 error (%): 23.8 -> 22.5 | Helped |
| Increase Tensor Sketch output dimension from 4K to 16K (ablation) | Tensor Sketch, CUB-200 · CUB-200, fine-tuned | Top-1 error (%): 23.2 -> 22.5 | Helped |
Closest papers by meaning
- Efficient Compact Bilinear Pooling via Kronecker Product (AAAI 2022 · similarity 0.92)
- Low-Rank Bilinear Pooling for Fine-Grained Classification (CVPR 2017 · similarity 0.91)
- I-Con: A Unifying Framework for Representation Learning (ICLR 2025 · similarity 0.83)
- Multilinear Operator Networks (ICLR 2024 · similarity 0.82)
- Few-Class Arena: A Benchmark for Efficient Selection of Vision Models and Dataset Difficulty Measurement (ICLR 2025 · similarity 0.82)
- 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.