Streaming Vlm
StreamingVLM: Real-Time Understanding for Infinite Video Streams
About
StreamingVLM enables real-time, stable understanding of effectively infinite video by keeping a compact KV cache and aligning training with streaming inference. It avoids quadratic cost and sliding-window pitfalls, runs up to 8 FPS on a single H100, and wins 66.18% vs GPT-4o mini on a new long-video benchmark. It also boosts general VQA without task-specific finetuning. You can grasp the gist by skimming this section first.
Open Source Health
- Stars
- 1,080
- Forks
- 67
- License
- MIT
- Last commit
- 1 years ago
Resources & Links
Related Categories
Vendor
MIT HAN Lab
Efficient AI Computing. PI: Song Han
Quick Links
Open Source
More by MIT HAN Lab
Related Products
Deep-Feature-Flow
Deep Feature Flow for Video Recognition
Top category match
MMAction
An open-source toolbox for action understanding based on PyTorch
Top category match
VideoPrism
Official repository for "VideoPrism: A Foundational Visual Encoder for Video Understanding" (ICML 2024)
Top category match
Temporal Segment Networks
Code & Models for Temporal Segment Networks (TSN) in ECCV 2016
Top category match
Cosmos Curator
Cosmos Curator is a powerful video curation system that processes, analyzes, and organizes video content using advanced AI models and distributed computing.
Top category match