MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition

ICCV 2023  ·  QiHao Zhao, Chen Jiang, Wei Hu, Fan Zhang, Jun Liu ·

Recently, multi-expert methods have led to significant improvements in long-tail recognition (LTR). We summarize two aspects that need further enhancement to contribute to LTR boosting: (1) More diverse experts; (2) Lower model variance. However, the previous methods didn't handle them well. To this end, we propose More Diverse experts with Consistency Self-distillation (MDCS) to bridge the gap left by earlier methods. Our MDCS approach consists of two core components: Diversity Loss (DL) and Consistency Self-distillation (CS). In detail, DL promotes diversity among experts by controlling their focus on different categories. To reduce the model variance, we employ KL divergence to distill the richer knowledge of weakly augmented instances for the experts' self-distillation. In particular, we design Confident Instance Sampling (CIS) to select the correctly classified instances for CS to avoid biased/noisy knowledge. In the analysis and ablation study, we demonstrate that our method compared with previous work can effectively increase the diversity of experts, significantly reduce the variance of the model, and improve recognition accuracy. Moreover, the roles of our DL and CS are mutually reinforcing and coupled: the diversity of experts benefits from the CS, and the CS cannot achieve remarkable results without the DL. Experiments show our MDCS outperforms the state-of-the-art by 1% $\sim$ 2% on five popular long-tailed benchmarks, including CIFAR10-LT, CIFAR100-LT, ImageNet-LT, Places-LT, and iNaturalist 2018. The code is available at https://github.com/fistyee/MDCS.

PDF Abstract ICCV 2023 PDF ICCV 2023 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Long-tail Learning CIFAR-100-LT (ρ=100) MDCS Error Rate 43.9 # 10
Long-tail Learning CIFAR-100-LT (ρ=50) MDCS Error Rate 39.9 # 9
Long-tail Learning CIFAR-10-LT (ρ=50) MDCS Error Rate 11.7 # 5
Long-tail Learning ImageNet-LT MDCS (ResNeXt-50) Top-1 Accuracy 61.8 # 11
Long-tail Learning iNaturalist 2018 MDCS(Resnet50) Top-1 Accuracy 75.6% # 9

Methods