DL-VTON: An Image-Based Virtual Try-On Framework Using Deep Learning Models

Virtual try-on (VTON) systems have emerged as transformative tools in fashion, enabling users to visualize how garments would appear on them without physical trials. This paper presents a novel 2D image-based virtual try-on pipeline that combines cutting-edge deep learning techniques to achieve photorealistic results. The proposed system incorporates OpenPose for human keypoint detection and body segmentation, Thin Plate Spline (TPS) deformation for precise garment alignment with body contours, and Contextual Attention for reconstructing exposed skin areas caused by garment misalignment. To ensure seamless integration of the garment with the person’s image, the pipeline employs the Vision Transformer (ViT), leveraging its advanced attention mechanisms for superior image blending and texture consistency. Designed to handle diverse body poses, garment styles, and intricate body-garment interactions, this framework is efficient, scalable, and accessible, making it ideal for applications in fashion e-commerce and virtual try-on experiences. By emphasizing simplicity and performance, this work sets a benchmark for the next generation of virtual try-on technologies.