Multi-Pose Virtual Try-on with Identity and Clothing Detail Preservation Using ResNet, VGG19 and MobileNet

This study presents a state-of-the-art high-resolution Virtual Try-On (VITON) system that uses deep learning and GANs to produce lifelike photos of individuals wearing specific clothing. In contrast to conventional techniques, our method uses a high-capacity conditional GAN framework, alignment-aware normalization, and sophisticated segmentation to improve visual quality. Important elements include an ALIAS generator to preserve texture details, U2-Net for accurate segmentation, and MediaPipe for real-time posture estimation. According to a comparison analysis, MobileNet is the best backbone in terms of both quality and efficiency. Superior realism is confirmed by evaluations on the VITON-HD dataset, which makes the system perfect for AR applications and online fashion shopping. Pose customization and texture refining are possible future developments.