Research map / Detection and segmentation / CNN features
Context-aware Attentional Pooling (CAP) for Fine-grained Visual Classification
Read the paperCheck 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 |
|---|---|---|---|
| Add CAP to a base CNN (ablation) | Fine-grained visual classification · Aircraft, Stanford Cars, Oxford-IIIT Pets | top-1 accuracy: 79.7 -> 88.8; 84.7 -> 91.5; 86.8 -> 94.7 | Helped |
| Add the classification module after CAP (ablation) | CAP with ResNet-50 on Aircraft · Aircraft | top-1 accuracy: 88.8 -> 94.9 | Helped |
| Add the classification module after CAP (ablation) | CAP with Xception · Stanford Cars | top-1 accuracy: 91.6 -> 95.7 | Helped |
| Add the classification module after CAP (ablation) | CAP with NASNet-Mobile · Oxford-IIIT Pets | top-1 accuracy: 95.6 -> 97.3 | Helped |
| Use 27 rather than 9 integral regions (ablation) | CAP with ResNet-50 · Aircraft | top-1 accuracy: 85.9 -> 94.9 | Helped |
| Increase the region count from 27 to 36 (ablation) | CAP with ResNet-50 · Aircraft | top-1 accuracy: 94.9 -> 91.2 | Hurt |
| Use CAP as the full method rather than the strongest cited primary-data baseline | Fine-grained visual classification · Food-101 | top-1 accuracy: 93.0 -> 98.6 | Helped |
| Use CAP rather than the strongest cited primary-data baseline | Fine-grained visual classification · NABirds | top-1 accuracy: 86.4 -> 91.0 | Helped |
| Use CAP rather than the strongest cited baseline | Fine-grained visual classification · Stanford Dogs | top-1 accuracy: 93.9 -> 96.1; 97.1 vs 96.1 | Mixed |
| Use CAP rather than the strongest cited baseline | Fine-grained visual classification · Oxford Flowers | top-1 accuracy: 96.4 -> 97.7; 97.7 vs 97.7 | Mixed |
Closest papers by meaning
- Efficient Representation Learning via Adaptive Context Pooling (ICML 2022 · similarity 0.91)
- Part-Stacked CNN for Fine-Grained Visual Categorization (CVPR 2016 · similarity 0.88)
- Contextual Convolutional Networks (ICLR 2023 · similarity 0.88)
- Learn to Pay Attention (ICLR 2018 · similarity 0.88)
- Attention to Scale: Scale-Aware Semantic Image Segmentation (CVPR 2016 · similarity 0.87)
- Revisiting the Integration of Convolution and Attention for Vision Backbone (NeurIPS 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.