Simultaneous Multi‐Attribute Image‐to‐Image Translation Using Parallel Latent Transform Networks

Abstract

Image-to-image translation has been widely studied. Since real-world images can often be described by multiple attributes, it is useful to manipulate them at the same time. However, most methods focus on transforming between two domains, and when they chain multiple single attribute transform networks together, the results are affected by the order of chaining, and the performance drops with the out-of-domain issue for intermediate results. Existing multi-domain transfer methods mostly manipulate multiple attributes by adding a list of attribute labels to the network feature, but they also suffer from interference of different attributes, and perform worse when multiple attributes are manipulated. We propose a novel approach to multiattribute image-to-image translation using several parallel latent transform networks, where multiple attributes are manipulated in parallel and simultaneously, which eliminates both issues. To avoid the interference of different attributes, we introduce a novel soft independence constraint for the changes caused by different attributes. Extensive experiments show that our method outperforms state-of-the-art methods. CCS Concepts • Computing methodologies → Image manipulation;

The Way to picture Reality: A Review, Evaluation and Recipe of Image Harmonization

“…The foreground is directly blended into the background by optimizing the Poisson blending loss and the style and content losses computed from the deep network, combined with iterative pixel updates using the L-BFGS solver to reconstruct blended regions [88]. Xu et al [89] utilized Latent transform networks (LTNs) to combine attributes while transforming features between images in parallel. Sbai et al [90] considered foreground category similarity when synthesizing pictures to control the authenticity of the overall photo, so as to achieve a more interesting effect of composite photos.…”

“…Compared with supervised deep learning, unsupervised deep learning appeared relatively earlier. Earlier appearance means more integration with Zhang et al [88]; Xu et al [89]; Sbai et al [90] traditional image information and methods. This type of method learns various information in the image through a large amount of data, and calculates the matching score within the image, thereby improving the matching degree between the background and the foreground.…”