Research map / Label-efficient and robust learning / Noisy labels
Attend in Groups: A Weakly-Supervised Deep Learning Framework for Learning From Web Data
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 |
|---|---|---|---|
| Add random grouping, attention, and the attention regularizer | average-pooling classifier · CompCars test set (WebCars) | accuracy: 66.86% -> 78.44% | Helped |
| Add random grouping to average pooling (ablation) | classifier without attention · CompCars test set (WebCars) | accuracy: 66.86% -> 69.83% | Helped |
| Add attention (ablation) | average-pooling classifier · CompCars test set (WebCars) | accuracy: 66.86% -> 73.64% | Helped |
| Add attention (ablation) | random-grouping classifier · CompCars test set (WebCars) | accuracy: 69.83% -> 76.58% | Helped |
| Add the attention regularizer (ablation) | AP+AT · CompCars test set (WebCars) | accuracy: 73.64% -> 70.77 % | Hurt |
| Add the attention regularizer (ablation) | RGT+AT · CompCars test set (WebCars) | accuracy: 76.58% -> 78.44% | Helped |
| Add attention to average pooling (ablation) | AP · CompCars test set, noise 60% | MAP: 74.42 -> 90.56 | Helped |
| Add group-wise training and the attention regularizer (ablation) | AP+AT · CompCars test set, noise 20% | MAP: 96.71 -> 98.12 | Helped |
| Increase group size from 2 to 4 (ablation) | RGT+AT+R · CompCars test set, noise 60% | MAP: 91.00 -> 91.98 | Helped |
| Increase group size from 2 to 4 (ablation) | RGT+AT+R · ILSVRC2012 validation set | accuracy: 71.24% -> 66.23% | Hurt |
| Add random grouping and the attention regularizer (ablation) | AP+AT · ILSVRC2012 validation set | accuracy: 67.68% -> 71.24% | Helped |
| Vary group size from 2 to 4 as noise increases (ablation) | RGT+AT+R · CompCars test set with noise levels 20%, 40%, and 60% | MAP: 98.12 (group size=2) -> 95.93 (group size=3) -> 91.98 (group size=4) | Mixed |
Closest papers by meaning
- Revisiting Weakly Supervised Pre-Training of Visual Perception Models (CVPR 2022 · similarity 0.88)
- Web-Supervised Network with Softly Update-Drop Training for Fine-Grained Visual Classification (AAAI 2020 · similarity 0.87)
- Weakly Supervised Image Classification Through Noise Regularization (CVPR 2019 · similarity 0.86)
- Exploiting weakly supervised visual patterns to learn from partial annotations (NeurIPS 2020 · similarity 0.86)
- Learning Facial Action Units From Web Images With Scalable Weakly Supervised Clustering (CVPR 2018 · similarity 0.86)
- Pre-Trained Vision-Language Models as Noisy Partial Annotators (AAAI 2025 · similarity 0.85)
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.