Article
Computer Science, Artificial Intelligence
Shuai Yang, Zhangyang Wang, Jiaying Liu, Zongming Guo
Summary: This paper introduces a novel controllable sketch-to-image translation framework for synthesizing and editing face images with hand-drawn sketches. The proposed method includes a dilation-based sketch refinement method and multi-level refinement for balancing between realism and structural consistency. The framework also explores advanced user controllability for fine-grained and semantic editing, demonstrated through extensive experiments showcasing visual quality and user control.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Jianyuan Sun, Hongchuan Yu, Jian J. Zhang, Junyu Dong, Hui Yu, Guoqiang Zhong
Summary: The study introduces a novel end-to-end generative adversarial fusion model, GAF, for generating more realistic color face images. By integrating an illumination distribution layer between two U-Net generators and introducing an attention mechanism, the image quality and face recognition accuracy are improved.
Article
Computer Science, Artificial Intelligence
Chunlei Peng, Congyu Zhang, Decheng Liu, Nannan Wang, Xinbo Gao
Summary: With the rapid development of generative adversarial networks, the synthesis of face photo-sketch has become increasingly important in law enforcement and entertainment. However, existing methods lack generalization ability and produce images with insufficient fidelity. In this study, we propose a novel cross-domain face photo-sketch synthesis framework called HiFiSketch, which can adjust the weights of generators for high-fidelity synthesis and manipulation, while preserving facial details effectively. Extensive experiments show that our method outperforms existing methods in terms of synthesis and manipulation of face photo-sketch.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Information Systems
B. Yogameena, Geeta Jakkamsetti, S. Aishwarya
Summary: This study proposes a method of sketch-face matching in videos using SpyGAN. By converting the detected faces in videos into sketches that focus on key facial regions and matching them using PCA-SIFT descriptors, high accuracy is achieved.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2022)
Article
Computer Science, Artificial Intelligence
Lan Yan, Wenbo Zheng, Chao Gou, Fei-Yue Wang
Summary: This paper proposes a novel Identity-sensitive Generative Adversarial Network (IsGAN) to address the problem of face photo-sketch synthesis by embedding identity information through adversarial learning. The model introduces identity recognition loss and cyclic-synthesized loss to preserve identifiable information and enforce structural consistency, achieving state-of-the-art performance on the CUFS and CUFSF datasets.
PATTERN RECOGNITION
(2021)
Article
Engineering, Electrical & Electronic
Xin Fang, Yiping Duan, Qiyuan Du, Xiaoming Tao, Fan Li
Summary: This study proposes a sketch-assisted face image coding method for human and machine vision using a joint training approach. The method introduces a new feature representation called color sketch and presents an end-to-end image codec framework with three models: image-to-image translation, coding, and reconstruction. The experimental results demonstrate significant bitrate savings on machine vision and comparable performance to state-of-the-art image coding methods on human vision.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Theory & Methods
Congyu Zhang, Decheng Liu, Chunlei Peng, Nannan Wang, Xinbo Gao
Summary: With the development of generative adversarial networks, face sketch synthesis is gaining attention for its promising prospects in entertainment and law enforcement. A novel generative adversarial network is proposed to synthesize sketches with similar shapes and rich details to photos. The method incorporates a cross-domain framework and a spatially adaptive denormalization module to improve the synthesis quality and details of the face images.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2022)
Article
Computer Science, Theory & Methods
Shuchao Duan, Zhenxue Chen, Q. M. Jonathan Wu, Lei Cai, Dan Lu
Summary: This paper proposes a novel method for face photo-sketch transformation, which addresses the traditional problems using a multi-scale gradients self-attention residual learning framework, and achieves both face photo-to-sketch synthesis and face sketch-to-photo synthesis simultaneously through a cycle framework.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2021)
Article
Computer Science, Artificial Intelligence
Osman Tursun, Simon Denman, Sridha Sridharan, Ethan Goan, Clinton Fookes
Summary: This paper proposes a simple and efficient framework for zero-shot sketch-based image retrieval (ZS-SBIR). It fine-tunes a pre-trained CNN and introduces multiple learning objects to learn discriminative, semantic, and domain invariant features. It also leverages pre-trained semantic knowledge using a semantic knowledge preservation loss. Extensive experiments show that the proposed method achieves state-of-the-art results on challenging datasets.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Software Engineering
Takato Yoshikawa, Yuki Endo, Yoshihiro Kanamori
Summary: In this paper, a sketch-based face image synthesis method is proposed, which generates diverse face images from a single sketch by using a three-stage framework of sketch refinement, detail enhancement, and appearance synthesis. The method achieves both realism and diversity in the output images.
Article
Computer Science, Information Systems
Fan Yang, Yang Wu, Zheng Wang, Xiang Li, Sakriani Sakti, Satoshi Nakamura
Summary: This paper proposes an Instance-level Heterogeneous Domain Adaptation (IHDA) framework to address domain adaptation issues in sketch-to-photo retrieval. By utilizing fine-tuning for identity label learning, shared attributes, and domain adaptation, instance-level knowledge transfer is facilitated.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Computer Science, Artificial Intelligence
Yinghui Zhang, Lejun Yu, Bo Sun, Jun He
Summary: In this study, an enhanced asymmetric CycleGAN framework (ENG-Face) is proposed to address the issues of mapping ambiguity and identity consistency in heterogeneous face synthesis. The experimental results demonstrate the effectiveness of ENG-Face in synthesizing heterogeneous face images in unpaired domains.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Theory & Methods
Ziming Yang, Jian Liang, Chaoyou Fu, Mandi Luo, Xiao-Yu Zhang
Summary: This article presents a new method for heterogeneous face recognition (FSIAD) that tackles the challenges of cross-domain discrepancy and facial attribute variation through identity-attribute disentanglement and face synthesis. Experimental results on multiple HFR databases validate the superior performance of FSIAD compared to previous approaches.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2022)
Article
Automation & Control Systems
Jun Yu, Xingxin Xu, Fei Gao, Shengjie Shi, Meng Wang, Dacheng Tao, Qingming Huang
Summary: Face photo-sketch synthesis remains challenging due to restrictions on structural realism and textural consistency. The proposed CA-GAN and SCA-GAN methods can generate visually comfortable, identity-preserving face sketches/photos with significant potential for generalization.
IEEE TRANSACTIONS ON CYBERNETICS
(2021)
Article
Computer Science, Artificial Intelligence
Lin Cao, Jianqiang Yin, Yanan Guo, Kangning Du, Fan Zhang
Summary: The authors propose a light semantic Transformer network to extract and model the semantic information of cross-modality images for sketch face recognition. They address the small-sample problem through meta-learning and introduce a hierarchical group linear transformation and parameter sharing to handle the complexity of the Transformer. They also propose a domain-adaptive focal loss to reduce cross-modality differences. Experimental results show improved recognition rates.
IET COMPUTER VISION
(2023)