StackMix: A complementary Mix algorithm

Abstract

Techniques combining multiple images as input/output have proven to be effective data augmentations for training convolutional neural networks. In this paper, we present StackMix: Each input is presented as a concatenation of two images, and the label is the mean of the two one-hot labels. On its own, StackMix rivals other widely used methods in the "Mix" line of work. More importantly, unlike previous work, significant gains across a variety of benchmarks are achieved by combining StackMix with existing Mix augmentation, effectively mixing more than two images. E.g., by combining StackMix with CutMix, test error in the supervised setting is improved across a variety of settings over CutMix, including 0.8\% on ImageNet, 3\% on Tiny ImageNet, 2\% on CIFAR-100, 0.5\% on CIFAR-10, and 1.5\% on STL-10. Similar results are achieved with Mixup.We further show that gains hold for robustness to common input corruptions and perturbations at varying severities with a 0.7\% improvement on CIFAR-100-C, by combining StackMix with AugMix over AugMix. On its own, improvements with StackMix hold across different number of labeled samples on CIFAR-100, maintaining approximately a 2\% gap in test accuracy -- down to using only 5\% of the whole dataset -- and is effective in the semi-supervised setting with a 2\% improvement with the standard benchmark $\Pi$-model. Finally, we perform an extensive ablation study to better understand the proposed algorithm.

References in corpus (23)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• A Simple Framework for Contrastive Learning of Visual Representations
• mixup: Beyond Empirical Risk Minimization
• Improved Regularization of Convolutional Neural Networks with Cutout
• Wide Residual Networks
• Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results
• Momentum Contrast for Unsupervised Visual Representation Learning
• MixMatch: A Holistic Approach to Semi-Supervised Learning
• Contrastive Multiview Coding
• Training Convolutional Networks with Noisy Labels
• Data Augmentation using Random Image Cropping and Patching for Deep CNNs
• Manifold Mixup: Better Representations by Interpolating Hidden States
• Big Self-Supervised Models are Strong Semi-Supervised Learners
• Unsupervised Visual Representation Learning by Context Prediction
• Realistic Evaluation of Deep Semi-Supervised Learning Algorithms
• Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules
• Adversarial Training Can Hurt Generalization
• Supervised Contrastive Learning
• Puzzle Mix: Exploiting Saliency and Local Statistics for Optimal Mixup
• Revisiting Self-Supervised Visual Representation Learning
• Time-Contrastive Networks: Self-Supervised Learning from Video
• Deep Pyramidal Residual Networks
• Diversity inducing Information Bottleneck in Model Ensembles