Lu9876

/

VideoTG_R1

like 0

This PR replaces the placeholder model card with detailed information for the VideoTG-R1 model.

It includes:

• Updating metadata with pipeline_tag: video-text-to-text, library_name: transformers, datasets: yeliudev/VideoMind-Dataset, and relevant tags like video-temporal-grounding, multimodal-llm, reinforcement-learning, and curriculum-learning.
• A link to the paper (VideoTG-R1: Boosting Video Temporal Grounding via Curriculum Reinforcement Learning on Reflected Boundary Annotations).
• A link to the GitHub repository (https://github.com/ldong1111/VideoTG-R1).
• A descriptive abstract and methodology visualizations.
• Comprehensive usage, training, and evaluation instructions with bash code snippets directly from the official GitHub README.

This update will significantly improve discoverability and usability for researchers interested in Video Temporal Grounding.