Joint Adaptive Representations for Image-Language Learning
This addresses the problem of computational inefficiency in image-language learning for researchers and practitioners, offering a more accessible approach, though it is incremental as it builds on existing fusion techniques.
The paper tackles the high computational cost and data requirements of vision-language models by proposing a joint adaptive representation learning method, achieving a 33% reduction in FLOPs and competitive performance with only 40M training examples and 39 GFLOPs, outperforming larger models.
Image-language learning has made unprecedented progress in visual understanding. These developments have come at high costs, as contemporary vision-language models require large model scales and amounts of data. We here propose a much easier recipe for image-language learning, which produces effective models, outperforming bigger and more expensive ones, often trained on orders of magnitude larger datasets. Our key finding is the joint learning of a compact vision and language representation, which adaptively and iteratively fuses the multi-modal features. This results in a more effective image-language learning, greatly lowering the FLOPs by combining and reducing the number of tokens for both text and images, e.g. a 33\% reduction in FLOPs is achieved, compared to baseline fusion techniques used by popular image-language models, while improving performance. This also allows the model to scale without a large increase in FLOPs or memory. In addition, we propose adaptive pre-training data sampling which improves the data efficiency. The proposed approach achieves competitive performance compared to much larger models, and does so with significantly less data and FLOPs. With only 40M training examples and with 39 GFLOPs our lightweight model outperforms many times larger state-of-the-art models of 2-20x more FLOPs and using bigger datasets some of which with close to 1B training examples.