openinterx

/

UGC-VideoCaptioner

like 4

Follow

Memories.ai Research 13

This PR significantly enhances the model card for UGC-VideoCaptioner-3B by:

• Adding pipeline_tag: video-text-to-text and library_name: transformers to the metadata for better discoverability and Hub integration.
• Including additional tags (multimodal, video-captioning, audio-visual, ugc) to provide more descriptive categories.
• Providing a detailed overview of the model based on its paper abstract.
• Adding direct links to the official Hugging Face paper page, the project's homepage, and the GitHub repository.
• Integrating visual elements such as the tiktok_qa_sample.png and benchmark.png from the repository.
• Presenting the model's "Benchmark Results" table for easy reference.
• Adding a "Quick Start" guide with a Python code example for local inference using the transformers library.
• Including "Evaluation" details and a BibTeX "Citation".

These additions make the model card more informative, user-friendly, and comprehensive on the Hugging Face Hub.