Virtual Try-On Preprocessing Pipeline

Advanced computer vision system for accurate clothing alignment

Tech Stack: Python OpenPose DensePose U-Net TensorFlow
Categories: Computer Vision Generative AI Image Processing
Note: This showcases my independent work on the preprocessing pipeline. The full virtual try-on system was developed as part of my graduation project with code maintained by my university team.

Project Impact

Developed a complete preprocessing pipeline that improved clothing alignment accuracy by 25% for virtual try-on systems, enabling more realistic fashion visualization.

Preprocessing Pipeline

Cloth Masking

U-Net based segmentation with adaptive padding for garment extraction

View Code

Body Parsing

Semantic segmentation for 12 body regions using HRNet

View Code

Pose Estimation

OpenPose 1.17 for 18-keypoint body pose detection

View Code

DensePose

Pixel-level body surface mapping using ResNet101-FPN

View Code

Technical Highlights

Input sample Output sample
View Full Repository

Developed as part of graduation research at Cairo University - Code available for preprocessing components only