Domain Generalization by Solving Jigsaw Puzzles

Human adaptability relies crucially on the ability to learn and merge knowledge both from supervised and unsupervised learning: the parents point out few important concepts, but then the children fill in the gaps on their own. This is particularly effective, because supervised learning can never be exhaustive and thus learning autonomously allows to discover invariances and regularities that help to generalize. In this paper we propose to apply a similar approach to the task of object recognition across domains: our model learns the semantic labels in a supervised fashion, and broadens its understanding of the data by learning from self-supervised signals how to solve a jigsaw puzzle on the same images. This secondary task helps the network to learn the concepts of spatial correlation while acting as a regularizer for the classification task. Multiple experiments on the PACS, VLCS, Office-Home and digits datasets confirm our intuition and show that this simple method outperforms previous domain generalization and adaptation solutions. An ablation study further illustrates the inner workings of our approach.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Classification Colored-MNIST(with spurious correlation) JiGen Accuracy 11.91 # 6
Domain Generalization NICO Animal JiGen (Resnet-18) Accuracy 84.95 # 3
Domain Generalization NICO Vehicle ResNet-18 Accuracy 77.39 # 4
Domain Generalization PACS JiGen (Alexnet) Average Accuracy 73.38 # 99
Domain Generalization PACS JiGen (Resnet-18) Average Accuracy 80.51 # 84
Domain Generalization PACS Deep All (Alexnet) Average Accuracy 71.52 # 106
Domain Generalization PACS Deep All (Resnet-18) Average Accuracy 79.05 # 88

Methods