Can I read VD-IT: Enhancing R-VOS with T2V Diffusion on EtoBox?

VD-IT: Enhancing R-VOS with T2V Diffusion by xepit98367 is a document available to read on EtoBox.

What is VD-IT: Enhancing R-VOS with T2V Diffusion about?

1. The document proposes a new framework called VD-IT for referring video object segmentation that leverages pre-trained text-to-video diffusion models. 2. VD-IT extracts video features from the diffusion model and uses a mask decoder to generate segmentation masks. It incorporates two key techniques: text-guided image projection to ensure feature consistency across time, and video-specific noise prediction to preserve feature fidelity. 3. Evaluation on four benchmarks shows VD-IT achieves competitive perf

Author

xepit98367

Language

EN