Image Captioners Are Scalable Vision Learners Too

Michael Tschannen, Manoj Kumar, Andreas Peter Steiner, Xiaohua Zhai, Neil Houlsby, Lucas Beyer

Keywords: contrastive learning, CLIP, CapPa, Cap, vision-language, image captioning, visual representation learning, weakly supervised learning, VLM, multimodal learning, VQA, image classification

TL;DR: We present an extensive comparison of contrastive representation learning and representation learning via image captioning from large image-text data sets.

Abstract: Contrastive pretraining on image-text pairs from the web is one of the most popular large-scale pretraining strategies for vision backbones, especially in the context of large multimodal models. At the same time, image captioning on this type of data is commonly considered an inferior pretraining strategy. In this paper, we perform a fair comparison of these two pretraining strategies, carefully matching training data, compute, and model capacity. Using a standard encoder-decoder transformer, we find that captioning alone is surprisingly effective: on classification tasks, captioning produces vision encoders competitive with contrastively pretrained encoders, while surpassing them on vision & language tasks. We further analyze the effect of the model architecture and scale, as well as the pretraining data on the representation quality, and find that captioning exhibits the same or better scaling behavior along these axes. Overall our results show that plain image captioning is a more powerful pretraining strategy than was previously believed. Code is available at [https://github.com/google-research/big_vision](https://github.com/google-research/big_vision).

Submission Number: 7713