Semantically Grounded QFormer for Efficient Vision Language Understanding (2311.07449v2)

Abstract: General purpose Vision LLMs (VLMs) have received tremendous interest in recent years, owing to their ability to learn rich vision-language correlations as well as their broad zero-shot competencies. One immensely popular line of work utilizes frozen unimodal models, by bridging vision representations to language using a trainable module called the QFormer. However, this method relies heavily on large-scale multimodal pretraining with huge computational overheads. To that end, we propose a more efficient framework for QFormer-based vision-language alignment. Our key idea relies on the observation that QFormer latents correspond more strongly to the frozen LLM's intermediate latent space. Consequently, instead of using QFormer latents as inputs to the LLM, we alter the framework by using the latents to directly condition the LLM latent space for image-to-text generation. We demonstrate the effectiveness of our approach against existing baselines in improving the efficiency of vision-language pretraining.