English

Solution for Large-scale Long-tailed Recognition with Noisy Labels

Computer Vision and Pattern Recognition 2021-06-22 v1 Artificial Intelligence

Abstract

This is a technical report for CVPR 2021 AliProducts Challenge. AliProducts Challenge is a competition proposed for studying the large-scale and fine-grained commodity image recognition problem encountered by worldleading ecommerce companies. The large-scale product recognition simultaneously meets the challenge of noisy annotations, imbalanced (long-tailed) data distribution and fine-grained classification. In our solution, we adopt stateof-the-art model architectures of both CNNs and Transformer, including ResNeSt, EfficientNetV2, and DeiT. We found that iterative data cleaning, classifier weight normalization, high-resolution finetuning, and test time augmentation are key components to improve the performance of training with the noisy and imbalanced dataset. Finally, we obtain 6.4365% mean class error rate in the leaderboard with our ensemble model.

Keywords

Cite

@article{arxiv.2106.10683,
  title  = {Solution for Large-scale Long-tailed Recognition with Noisy Labels},
  author = {Yuqiao Xian and Jia-Xin Zhuang and Fufu Yu},
  journal= {arXiv preprint arXiv:2106.10683},
  year   = {2021}
}

Comments

3 pages

R2 v1 2026-06-24T03:23:57.640Z