EVTAR: End-to-End Try on with Additional Unpaired Visual Reference

Liuzhuozheng Li, Yue Gong, Shanyuan Liu, Liebucha Wu, Zanyi Wang, Dengyang Jiang, Bo Cheng, YuhangMa, Dawei Leng, Yuhui Yin

Keywords: Virtual Tryon

Abstract: We propose EVTER, an end-to-end virtual try-on model that incorporates additional reference images. Most existing virtual try-on models rely on complex inputs, such as images with agnostic clothing areas, human pose, densepose, and body keypoints, making them labor-intensive and impractical for real-world applications. In contrast, EVTER addresses these challenges by adopting an end-to-end training strategy, allowing for simple inference with only the source image and target clothing as inputs. The model generates try-on images without the need for masking. Moreover, to enhance try-on quality, our model can utilize additional reference images, inspired by how humans typically select clothing. To enable this capability, we built a dataset with supplementary reference images for training. We evaluate our model on popular benchmarks, and the results validate the effectiveness of our proposed approach.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 4403