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.
翻译:这是CVPR 2021 AliProduces Challenge的技术报告。 AliProduces Challenge是研究世界领先的电子商务公司遇到的大规模和精细的商品形象识别问题的一种建议竞争。大型产品识别同时应对了噪音说明、不平衡(长尾)数据分布和精细分类的挑战。在我们的解决方案中,我们采用了CNNs和变异器的最新模型结构,包括ResNest、高效NetV2和DeiT。我们发现迭代数据清理、分类器重量正常化、高分辨率微调和测试时间增强是改进噪音和不平衡数据集培训业绩的关键组成部分。最后,我们用共同模型在领导板上获得了6.4365 % 的平均班级差率。