Sketch2RealGAN: A Conditional GAN-based Method for Generating Clothing Images from Sketches
S.C. Zhang, Haopeng Lei, He Zhang · 2023
This algorithm is an innovative algorithm that can automatically generate clothing images from clothing sketches. In order to improve the generation performance, the algorithm uses feature extraction network to extract semantic information from the original sketch, and uses semantic feature encoder to encode these semantic information into feature tensors. High quality clothing image generation is realized by input both the original sketch and the feature tensor into the conditional generation adversarial network. In addition, a two-stage generation algorithm is proposed to generate clothing images from original sketches. The algorithm uses the same model structure in both stages, and finally realizes the clothing image generation by gradually generating the intermediate image.