TG-Net: Reconstruct visual wood texture with semantic attention. (February 2022)

• Record Type: Journal Article
• Authors: Chen, Jiahao; Ge, Yilin; Wang, Quan; Zhang, Yizhuo
• Is Part Of: Computers & graphics. Volume 102(2022)
• Page Start: 546
• Page End: 553
• Publication Date: 2022-02
• Subjects: Wood defect treatment -- Image inpainting -- Generative adversarial networks -- Attention mechanism; Computer graphics -- Periodicals; 006.6
• DOI: 10.1016/j.cag.2021.09.006
• Languages: English
• ISSNs: 0097-8493

Abstract

Abstract: Defects, which arise during the growth and processing of the wood, affect the visual and mechanical properties. To improve the utilization of wood with defects, we propose an image inpainting method that reconstructs the texture at defects. The result of image inpainting is visually very consistent with the surrounding healthy board. According to the result, as an application, the plate that is most similar to the surrounding texture could be matched to fill the missing position due to the removal of defects. First, the image inpainting is divided into two stages: coarse inpainting and detailed inpainting. The coarse generator quickly generates the prediction result of the missing area, and the result is fed to the detailed generator to produce a more refined image. Then, the skip connection is applied to the classical Encoder–Decoder in order to solve the problem of information loss due to upsampling. Next, the attention block is introduced in the detailed generator to pay more attention to the meaningful semantics. Finally, normalizing missing and known regions together would result in a bias in the mean and standard deviation, so Partition Normalization was used instead of Batch Normalization. The feature map is divided into two parts, foreground and background, and the two parts are normalized separately. The experimental results show that the results of image inpainting are still high precision with less time spent on the training network. Code, demo and models are available at: https://github.com/NEFUJoeyChen/TG-Net . Graphical abstract: Highlights: The network design of coarse-to-fine makes the model generate more fine texture. This is further strengthened by the attention mechanism in the detailed inpainting stage. The skip connection method is adopted in the Encoder–Decoder, which connects the high-level with the low-level semantics. It restores the details in the image. Partition normalization is different from the previous normalization method. The known region and the masked unknown region in the image are normalized respectively. It deals with the inconsistent mean and standard deviation.