Towards Calibrated Model for Long-Tailed Visual Recognition from Prior Perspective

NeurIPS 2021  ·  Zhengzhuo Xu, Zenghao Chai, Chun Yuan ·

Real-world data universally confronts a severe class-imbalance problem and exhibits a long-tailed distribution, i.e., most labels are associated with limited instances. The na\"ive models supervised by such datasets would prefer dominant labels, encounter a serious generalization challenge and become poorly calibrated. We propose two novel methods from the prior perspective to alleviate this dilemma. First, we deduce a balance-oriented data augmentation named Uniform Mixup (UniMix) to promote mixup in long-tailed scenarios, which adopts advanced mixing factor and sampler in favor of the minority. Second, motivated by the Bayesian theory, we figure out the Bayes Bias (Bayias), an inherent bias caused by the inconsistency of prior, and compensate it as a modification on standard cross-entropy loss. We further prove that both the proposed methods ensure the classification calibration theoretically and empirically. Extensive experiments verify that our strategies contribute to a better-calibrated model, and their combination achieves state-of-the-art performance on CIFAR-LT, ImageNet-LT, and iNaturalist 2018.

PDF Abstract NeurIPS 2021 PDF NeurIPS 2021 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Long-tail Learning CIFAR-100-LT (ρ=10) UniMix+Bayias (ResNet-32) Error Rate 38.75 # 24
Long-tail Learning CIFAR-100-LT (ρ=100) Prior-LT Error Rate 53.59 # 40
Long-tail Learning CIFAR-100-LT (ρ=100) UniMix+Bayias (ResNet-32) Error Rate 54.55 # 44
Long-tail Learning CIFAR-10-LT (ρ=10) Prior-LT Error Rate 12.20 # 42
Long-tail Learning CIFAR-10-LT (ρ=10) UniMix+Bayias Error Rate 10.34 # 26

Methods