Welcome to OpenGVLab! We are a research group from Shanghai AI Lab focused on Vision-Centric AI research. The GV in our name, OpenGVLab, means general vision, a general understanding of vision, so little effort is needed to adapt to new vision-based tasks.

Models

• InternVL: a pioneering open-source alternative to GPT-4V.
• InternImage: a large-scale vision foundation models with deformable convolutions.
• InternVideo: large-scale video foundation models for multimodal understanding.
• VideoChat: an end-to-end chat assistant for video comprehension.
• All-Seeing-Project: towards panoptic visual recognition and understanding of the open world.

Datasets

• ShareGPT4o: a groundbreaking large-scale resource that we plan to open-source with 200K meticulously annotated images, 10K videos with highly descriptive captions, and 10K audio files with detailed descriptions.
• InternVid: a large-scale video-text dataset for multimodal understanding and generation.
• MMPR: a high-quality, large-scale multimodal preference dataset.

Benchmarks

• MVBench: a comprehensive benchmark for multimodal video understanding.
• CRPE: a benchmark covering all elements of the relation triplets (subject, predicate, object), providing a systematic platform for the evaluation of relation comprehension ability.
• MM-NIAH: a comprehensive benchmark for long multimodal documents comprehension.
• GMAI-MMBench: a comprehensive multimodal evaluation benchmark towards general medical AI.