EVA: Exploring the Limits of Masked Visual Representation Learning at Scale

TL;DR

EVA is a large-scale vision foundation model trained solely on publicly available data, achieving state-of-the-art results across various vision tasks and serving as a bridge to multi-modal models, with scalable architecture and improved training stability.

Contribution

The paper introduces EVA, a scalable vision foundation model trained with a simple masked image modeling task, setting new benchmarks and enabling efficient multi-modal training.

Findings

EVA achieves top performance on image recognition, detection, and segmentation tasks.

Scaling EVA to one billion parameters improves transfer learning capabilities.

Initializing CLIP with EVA enhances training stability and reduces data requirements.

Abstract

We launch EVA, a vision-centric foundation model to explore the limits of visual representation at scale using only publicly accessible data. EVA is a vanilla ViT pre-trained to reconstruct the masked out image-text aligned vision features conditioned on visible image patches. Via this pretext task, we can efficiently scale up EVA to one billion parameters, and sets new records on a broad range of representative vision downstream tasks, such as image recognition, video action recognition, object detection, instance segmentation and semantic segmentation without heavy supervised training. Moreover, we observe quantitative changes in scaling EVA result in qualitative changes in transfer learning performance that are not present in other models. For instance, EVA takes a great leap in the challenging large vocabulary instance segmentation task: our model achieves almost the same state-of-the-art performance on LVISv1.0 dataset with over a thousand categories and COCO dataset with only eighty categories. Beyond a pure vision encoder, EVA can also serve as a vision-centric, multi-modal pivot to connect images and text. We find initializing the vision tower of a giant CLIP from EVA can greatly stabilize the training and outperform the training from scratch counterpart with much fewer samples and less compute, providing a new direction for scaling up and accelerating the costly training of multi-modal foundation models. To facilitate future research, we release all the code and models at https://github.com/baaivision/EVA.