- Optimized the Fabric Try-On feature in the Style3D iCreate product. Preprocessed clothing images using PyTorch, OpenCV, and PIL, including converting images to grayscale, normalizing pixel value distributions, and enhancing the brightness, contrast
- Developed seamless fabric image tiling algorithms and used Photoshop layer blending algorithm to overlay fabric patterns onto
- Employed Inpainting and Outpainting techniques in Stable Diffusion to improve fabric images, ensuring seamless fabric tiling by filling in missing areas and extending the image edges
- Conducted in-depth research on multi-modal models, comparing the performance of InternVL, IXC-2.5, and Florence-2 in various applications. Enhanced Stable Diffusion's performance in Chinese language and traditional cultural content generation by
- Delivered presentations on multi-modal algorithms, sharing research findings and technical applications with the AI team to boost technical capabilities and innovation