jroessler/flux-kontext-segmentation-socks

Socks-LoRA for Segmenting Long Sock Regions in Fashion Images

Model Overview

Socks-LoRA is a lightweight LoRA adapter designed for the FLUX.1 Kontext generative model, enabling automatic segmentation of the "long socks" region on fashion images—even when long socks are absent from the original image. This LoRA is specifically intended for virtual try-on and inpainting workflows, where constructing a precise mask for new or modified clothing items is critical.

Intended Use

• Automatically generating a segmentation mask for the "long sock" area, including lower legs and feet, in fashion/on-model images.
• Facilitating downstream virtual try-on, inpainting, and style transfer tasks by providing a consistent mask for input images. Prompts used:
• Caption: put a green mask for socks and lower leg skin on the person

Acknowledgments

Based on the FLUX.1 Kontext generative model, trained with Ostris’ AI Toolkit. See video instructions.