Scaling Language-Free Visual Representation Learning
This work addresses the problem of improving vision-centric representation learning for AI researchers by showing that visual SSL can match language-supervised methods at scale, though it is incremental as it builds on existing SSL and CLIP approaches.
The study tackled the performance gap between visual self-supervised learning (SSL) and Contrastive Language-Image Pretraining (CLIP) in multimodal tasks like Visual Question Answering (VQA) by training both on the same MetaCLIP data, finding that visual SSL scales better with data and model capacity, achieving CLIP-level performance on VQA and classic vision benchmarks without language supervision.
Visual Self-Supervised Learning (SSL) currently underperforms Contrastive Language-Image Pretraining (CLIP) in multimodal settings such as Visual Question Answering (VQA). This multimodal gap is often attributed to the semantics introduced by language supervision, even though visual SSL and CLIP models are often trained on different data. In this work, we ask the question: "Do visual self-supervised approaches lag behind CLIP due to the lack of language supervision, or differences in the training data?" We study this question by training both visual SSL and CLIP models on the same MetaCLIP data, and leveraging VQA as a diverse testbed for vision encoders. In this controlled setup, visual SSL models scale better than CLIP models in terms of data and model capacity, and visual SSL performance does not saturate even after scaling up to 7B parameters. Consequently, we observe visual SSL methods achieve CLIP-level performance on a wide range of VQA and classic vision benchmarks. These findings demonstrate that pure visual SSL can match language-supervised visual pretraining at scale, opening new opportunities for vision-centric representation learning.