This PR adds a comprehensive model card for ReChannel, linking it to the paper [From RGB Generation to Dense Field Readout: Pixel-Space Dense Prediction with Text-to-Image Models](https://huggingface.co/papers/2607.06553) and its GitHub repository. It also populates the YAML metadata with the `image-to-image` pipeline tag and the `mit` license.
Upload 4b/refseg/lora_tail.safetensors with huggingface_hub
