Ricky06662

/

VisionReasoner-7B

like 7

This PR significantly improves the model card for Ricky06662/VisionReasoner-7B by:

• Updating the pipeline_tag to image-segmentation to accurately reflect the model's core task of generating pixel-level masks, enhancing its discoverability on the Hugging Face Hub.
• Clarifying paper references by highlighting its foundation in the primary paper, "Seg-Zero: Reasoning-Chain Guided Segmentation via Cognitive Reinforcement", while also linking to "VisionReasoner: Unified Visual Perception and Reasoning via Reinforcement Learning".
• Correcting code and project page links to the official Seg-Zero GitHub repository: https://github.com/dvlab-research/Seg-Zero.
• Expanding the model description with detailed insights from the paper abstract and GitHub README, covering its decoupled architecture, reinforcement learning training, zero-shot generalization capabilities, and performance highlights.
• Adding additional datasets (Ricky06662/refCOCOg_9k_840, Ricky06662/VisionReasoner_multi_object_7k_840) to the metadata for better completeness.
• Embedding key visuals (overview, pipeline, examples) directly from the GitHub README to provide a richer visual context.
• Including a comprehensive citation section with BibTeX entries for both associated papers.

These changes ensure the model card is more informative, accurate, and discoverable for researchers and users.