LLaVA-OneVision: Easy Visual Task Transfer

TL;DR

LLaVA-OneVision is a versatile large multimodal model that excels across single-image, multi-image, and video tasks, demonstrating strong transfer learning and emerging capabilities in visual understanding.

Contribution

It introduces a unified model capable of handling diverse visual scenarios with effective transfer learning, advancing open large multimodal models.

Findings

First single model to excel in image, multi-image, and video tasks

Demonstrates strong transfer learning across modalities

Achieves new capabilities in video understanding

Abstract

We present LLaVA-OneVision, a family of open large multimodal models (LMMs) developed by consolidating our insights into data, models, and visual representations in the LLaVA-NeXT blog series. Our experimental results demonstrate that LLaVA-OneVision is the first single model that can simultaneously push the performance boundaries of open LMMs in three important computer vision scenarios: single-image, multi-image, and video scenarios. Importantly, the design of LLaVA-OneVision allows strong transfer learning across different modalities/scenarios, yielding new emerging capabilities. In particular, strong video understanding and cross-scenario capabilities are demonstrated through task transfer from images to videos.