Data-Efficient Image Recognition with Contrastive Predictive Coding

ICML 2020

Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this `semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10% Top-5 accuracy and supervised methods by 20%. Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. Speakers: Olivier Henaff, Aaron van den Oord, Ali Eslami, Ali Razavi, Aravind Srinivas, Carl Doersch, Jeffrey De Fauw