Data-Efficient Language-Supervised Zero-Shot Learning with Self-Distillation

TL;DR

This paper introduces a data-efficient zero-shot learning method using self-distillation that significantly reduces data requirements while achieving state-of-the-art results in image classification tasks.

Contribution

It proposes a contrastive distillation approach that learns from noisy image-text pairs using soft labels, requiring 133 times less data than CLIP.

Findings

Achieves 73% relative improvement on ImageNet 21k+1k zero-shot classification.

Outperforms CLIP by 10.5% on Google Open Images zero-shot evaluation.

Uses only 3M image-text pairs, significantly less data than previous methods.

Abstract

Traditional computer vision models are trained to predict a fixed set of predefined categories. Recently, natural language has been shown to be a broader and richer source of supervision that provides finer descriptions to visual concepts than supervised "gold" labels. Previous works, such as CLIP, use a simple pretraining task of predicting the pairings between images and text captions. CLIP, however, is data hungry and requires more than 400M image text pairs for training. We propose a data-efficient contrastive distillation method that uses soft labels to learn from noisy image-text pairs. Our model transfers knowledge from pretrained image and sentence encoders and achieves strong performance with only 3M image text pairs, 133x smaller than CLIP. Our method exceeds the previous SoTA of general zero-shot learning on ImageNet 21k+1k by 73% relatively with a ResNet50 image encoder and DeCLUTR text encoder. We also beat CLIP by 10.5% relatively on zero-shot evaluation on Google Open Images (19,958 classes).