Research map / Label-efficient and robust learning / Recognition and long-tail
Web-Supervised Network with Softly Update-Drop Training 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 |
|---|---|---|---|
| use probability cross-entropy global sample dropping instead of Co-teaching | VGG-16 web-supervised · CUB200 | ACA (%): 73.85 -> 77.22 | Helped |
| same, vs Co-teaching | VGG-16 web-supervised · Cars-196 | ACA (%): 73.10 -> 78.71 | Helped |
| same, vs Co-teaching | VGG-16 web-supervised · FGVC-Aircraft | ACA (%): 72.76 -> 72.88 | No clear effect |
| replace VGG-16 with VGG-19 (ablation) | Softly update-drop · CUB200 | ACA (%): 77.22 -> 75.87 | Hurt |
| replace VGG-16 with ResNet-34 (ablation) | Softly update-drop · CUB200 | ACA (%): 77.22 -> 74.99 | Hurt |
| increase web images per class from 50 to 100 (ablation) | Softly update-drop · CUB200 | ACA (%): 71.87 -> 77.22 | Helped |
| use single network instead of two-network Co-teaching framework (ablation) | Softly update-drop · CUB200 | ACA (%): 75.46 -> 77.22 | Helped |
| raise maximum drop rate tau from 0.25 to 0.35 (ablation) | Softly update-drop · CUB200 | ACA (%) | Hurt |
| vary tau 0.15-0.3 and Tk 5-20 (ablation) | Softly update-drop · CUB200 | ACA (%) | No clear effect |
| increase Tk to 25 (ablation) | Softly update-drop · CUB200 | ACA (%) | Hurt |
| global sample selection instead of mini-batch selection (ablation) | Softly update-drop · CUB200 | selection stability (overlap rate) | Helped |
Closest papers by meaning
- Webly Supervised Learning Meets Zero-Shot Learning: A Hybrid Approach for Fine-Grained Classification (CVPR 2018 · similarity 0.88)
- Attend in Groups: A Weakly-Supervised Deep Learning Framework for Learning From Web Data (CVPR 2017 · similarity 0.87)
- Learning from Web Data Using Adversarial Discriminative Neural Networks for Fine-Grained Classification (AAAI 2019 · similarity 0.87)
- Advancing Fine-Grained Classification by Structure and Subject Preserving Augmentation (NeurIPS 2024 · similarity 0.84)
- Fine-grained Classes and How to Find Them (ICML 2024 · similarity 0.83)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.83)
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.