Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Low-Shot Visual Recognition by Shrinking and Hallucinating Features
Bharath Hariharan, Ross Girshick
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 squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · ImageNet unseen classes, 10 per class | top-1 accuracy: 29.79 -> 30.83 | Improved |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · ImageNet unseen classes, 2 per class | top-5 accuracy: 34.45 -> 38.76 | Improved |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · ImageNet unseen classes, 20 per class | top-1 accuracy: 35.75 -> 35.10 | Hurt |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · ImageNet unseen classes, all examples | top-1 accuracy: 51.41 -> 50.14 | Hurt |
| train with Siamese loss instead of gradient-norm loss | 10-layer ResNet · ImageNet unseen classes, 1 per class | top-1 accuracy: 9.80 -> 7.20 | Hurt |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · Birds, 1 per class | top-1 accuracy: 11.35 -> 13.50 | Improved |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · Aircraft, 5 per class | top-1 accuracy: 16.78 -> 16.76 | No clear effect |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · Aircraft, all examples | top-1 accuracy: 37.95 -> 35.07 | Hurt |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · Oxford flowers, 1 per class | top-1 accuracy: 42.95 -> 38.90 | Hurt |
| add squared gradient-norm penalty on small-batch classifier loss | 10-layer ResNet · VOC 2007 classification, 5 per class | accuracy: 51.94 -> 56.33 | Improved |
Closest papers by meaning
- Improving Intervention Efficacy via Concept Realignment in Concept Bottleneck Models (ECCV 2024 · similarity 0.87)
- AdaDiff: Accelerating Diffusion Models through Step-Wise Adaptive Computation (ECCV 2024 · similarity 0.86)
- Hallucinating IDT Descriptors and I3D Optical Flow Features for Action Recognition With CNNs (ICCV 2019 · similarity 0.85)
- What Does a Platypus Look Like? Generating Customized Prompts for Zero-Shot Image Classification (ICCV 2023 · similarity 0.85)
- Emerging Properties in Self-Supervised Vision Transformers (ICCV 2021 · similarity 0.84)
- Dynamic Perceiver for Efficient Visual Recognition (ICCV 2023 · similarity 0.84)
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.
