| Improving Image Recognition by Retrieving from Web-Scale Image-Text Data | | 83.6 | | | | MAM (ViT-B/16) | 2023-04-11 |
| CurriculumNet: Weakly Supervised Learning from Large-Scale Web Images | ✓ Link | 79.3% | 93.6% | | | CurriculumNet (InceptionResNet-v2) | 2018-08-03 |
| CMW-Net: Learning a Class-Aware Sample Weighting Mapping for Robust Deep Learning | ✓ Link | 77.9% | 92.6% | | | ERM + CMW-Net-SL (InceptionResNet-V2) | 2022-02-11 |
| Learning with Neighbor Consistency for Noisy Labels | ✓ Link | 76.8 | | | | NCR+Mixup+DA (ResNet-50) | 2022-02-04 |
| Correlated Input-Dependent Label Noise in Large-Scale Image Classification | | 76.6% | 92.1% | | | Heteroscedastic (InceptionResNet-v2) | 2021-05-19 |
| Multi-label Iterated Learning for Image Classification with Label Ambiguity | | 76.5% | 90.9% | 68.7 | 86.4 | MILe (ResNet-50-D) | 2021-11-23 |
| Webly Supervised Image Classification with Self-Contained Confidence | ✓ Link | 75.78% | 91.07% | 70.66% | 88.46% | SCC (ResNet50-D) | 2020-08-27 |
| Learning with Neighbor Consistency for Noisy Labels | ✓ Link | 75.7% | | | | NCR (ResNet-50) | 2022-02-04 |
| Webly Supervised Image Classification with Metadata: Automatic Noisy Label Correction via Visual-Semantic Graph | ✓ Link | 75.48% | 90.15% | 69.42% | 87.29% | VSGraph-LC (ResNet50) | 2020-10-12 |
| Multi-label Iterated Learning for Image Classification with Label Ambiguity | | 75.2% | 90.3% | 67.1 | 85.6 | MILe (ResNet-50) | 2021-11-23 |
| Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization | ✓ Link | 75.0% | 90.6% | 67.1% | 86.7% | HAR (InceptionResNet-v2) | 2020-06-29 |
| Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels | ✓ Link | 74.3% | 90.5% | 67.5% | 87.2% | MentorMix (InceptionResNet-V2) | 2019-11-21 |
| MoPro: Webly Supervised Learning with Momentum Prototypes | ✓ Link | 73.9% | 90.0% | 67.8% | 87.0% | MoPro (ResNet-50) | 2020-09-17 |
| Learning from Web Data with Self-Organizing Memory Module | | 72.2% | 89.5% | 65.0% | 85.1% | SOMNet (ResNet-50) | 2019-06-28 |
| CurriculumNet: Weakly Supervised Learning from Large-Scale Web Images | ✓ Link | 72.1% | 89.2% | 64.8% | 84.9% | CurriculumNet (Inception-v2） | 2018-08-03 |
| MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels | ✓ Link | 70.8% | 88.0% | 62.5% | 83.0% | MentorNet (InceptionResNet-V2) | 2017-12-14 |