VLMAE: Vision-Language Masked Autoencoder
This work addresses focal bias in vision-language pre-training, which is a domain-specific problem for researchers and practitioners in multi-modal AI, and it is incremental as it builds on existing masked autoencoder methods.
The paper tackles the problem of focal bias in vision-language pre-training by proposing VLMAE, a vision-language masked autoencoder framework that uses visual generative learning to acquire fine-grained and unbiased features, achieving better performance in tasks like visual question answering and image-text retrieval with up to 20% pre-training speedup.
Image and language modeling is of crucial importance for vision-language pre-training (VLP), which aims to learn multi-modal representations from large-scale paired image-text data. However, we observe that most existing VLP methods focus on modeling the interactions between image and text features while neglecting the information disparity between image and text, thus suffering from focal bias. To address this problem, we propose a vision-language masked autoencoder framework (VLMAE). VLMAE employs visual generative learning, facilitating the model to acquire fine-grained and unbiased features. Unlike the previous works, VLMAE pays attention to almost all critical patches in an image, providing more comprehensive understanding. Extensive experiments demonstrate that VLMAE achieves better performance in various vision-language downstream tasks, including visual question answering, image-text retrieval and visual grounding, even with up to 20% pre-training speedup.