Search

Your search keyword '"Sun, Siyu"' showing total 1,155 results
Start Over Author "Sun, Siyu"
1,155 results on '"Sun, Siyu"'

2. Rethinking Classifier Re-Training in Long-Tailed Recognition: A Simple Logits Retargeting Approach

Author

Lu, Han, Sun, Siyu, Xie, Yichen, Zhang, Liqing, Yang, Xiaokang, and Yan, Junchi

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence
Abstract

In the long-tailed recognition field, the Decoupled Training paradigm has demonstrated remarkable capabilities among various methods. This paradigm decouples the training process into separate representation learning and classifier re-training. Previous works have attempted to improve both stages simultaneously, making it difficult to isolate the effect of classifier re-training. Furthermore, recent empirical studies have demonstrated that simple regularization can yield strong feature representations, emphasizing the need to reassess existing classifier re-training methods. In this study, we revisit classifier re-training methods based on a unified feature representation and re-evaluate their performances. We propose a new metric called Logits Magnitude as a superior measure of model performance, replacing the commonly used Weight Norm. However, since it is hard to directly optimize the new metric during training, we introduce a suitable approximate invariant called Regularized Standard Deviation. Based on the two newly proposed metrics, we prove that reducing the absolute value of Logits Magnitude when it is nearly balanced can effectively decrease errors and disturbances during training, leading to better model performance. Motivated by these findings, we develop a simple logits retargeting approach (LORT) without the requirement of prior knowledge of the number of samples per class. LORT divides the original one-hot label into small true label probabilities and large negative label probabilities distributed across each class. Our method achieves state-of-the-art performance on various imbalanced datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist2018.

Published
2024

3. Taming the Power of Diffusion Models for High-Quality Virtual Try-On with Appearance Flow

Author

Gou, Junhong, Sun, Siyu, Zhang, Jianfu, Si, Jianlou, Qian, Chen, and Zhang, Liqing

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Virtual try-on is a critical image synthesis task that aims to transfer clothes from one image to another while preserving the details of both humans and clothes. While many existing methods rely on Generative Adversarial Networks (GANs) to achieve this, flaws can still occur, particularly at high resolutions. Recently, the diffusion model has emerged as a promising alternative for generating high-quality images in various applications. However, simply using clothes as a condition for guiding the diffusion model to inpaint is insufficient to maintain the details of the clothes. To overcome this challenge, we propose an exemplar-based inpainting approach that leverages a warping module to guide the diffusion model's generation effectively. The warping module performs initial processing on the clothes, which helps to preserve the local details of the clothes. We then combine the warped clothes with clothes-agnostic person image and add noise as the input of diffusion model. Additionally, the warped clothes is used as local conditions for each denoising process to ensure that the resulting output retains as much detail as possible. Our approach, namely Diffusion-based Conditional Inpainting for Virtual Try-ON (DCI-VTON), effectively utilizes the power of the diffusion model, and the incorporation of the warping module helps to produce high-quality and realistic virtual try-on results. Experimental results on VITON-HD demonstrate the effectiveness and superiority of our method., Comment: Accepted by ACMMM 2023

Published
2023
Full Text
View/download PDF

10. Cognitive Diagnostic Analysis of Students' Mathematical Competency Based on the DINA Model

Author

Xu, Tianshu, Wu, Xiaopeng, Sun, Siyu, and Kong, Qiping

Abstract

Considering the importance of mathematics in modern society, it is crucial to understand the cognitive processes involved in the acquisition of complex mathematical competency. As a new generation of evaluation theory, cognitive diagnosis has its unique advantages in personalized evaluation. Based on the mathematical cognitive framework of Trends in International Mathematics and Science Study (TIMSS)-2011 and the Chinese mathematics curriculum, this research has formed a mathematical competency model composed of seven cognitive attributes. Sixty-seven released mathematical items for the fourth grade in TIMSS-2011 were used as assessment tools in this study. The deterministic inputs, noisy, "and" gate model was selected as the cognitive diagnosis model in this research, and the parameters of the model were evaluated according to the response data, based on which the effectiveness of the assessment tool was further verified, forming the evaluation framework of students' mathematical competency. This framework was used to analyze the data of 573 students' mathematical competency from Shanghai, China, specifically from three aspects: attribute mastery probability, learning path, and knowledge structure. Results show that students' performance in mastering attributes of mathematical cognition is excellent on the whole; some students' learning paths are leapfrog; there are certain differences in students' knowledge structure despite that they have the same total score. This research is performed as a systematic case study of the evaluation of students' mathematical competency and also provides a new perspective in assessing other knowledge and skills.

Published
2023
Full Text
View/download PDF

12. Two-Stage Enhancement for Recommendation Systems Based on Contrastive Learning

Author

Sun, Siyu, Cai, Tianyu, Yan, Fanli, Ju, Shenggen, Goos, Gerhard, Series Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Jin, Cheqing, editor, Yang, Shiyu, editor, Shang, Xuequn, editor, Wang, Haofen, editor, and Zhang, Yong, editor

Published
2024
Full Text
View/download PDF

15. Changes of the exotropia deviation in children with basic type of intermittent exotropia before and after monocular occlusion test

Author

Liu Yurong, Liu Yangzi, Sun Siyu, and Wang Lijing

Subjects
intermittent exotropia, monocular occlusion test, exotropia deviation, Ophthalmology, RE1-994
Abstract

AIM: To compare the changes of exotropia deviation in children with basic type of intermittent exotropia before and after monocular occlusion test.METHODS: Prospective clinical study. A total of 258 children with basic type of intermittent exotropia who underwent strabismus correction in our hospital from July 2021 to September 2022 were selected, including 122 males and 136 females, aged from 5 to 12 years, with an average age of 8.0±3.1 years. The exotropia deviation was measured at distance(6 m)and near(33 cm)fixation by prism alternating occlusion method, and the exotropia deviation was checked again after covering the non-dominant eyes of children for 40 min.RESULTS: Before and after the monocular occlusion test, the exotropia deviation at distance(6 m)fixation was 28.23△±10.79△ and 29.79△±10.85△, respectively(t=-0.903, P=0.368), while the exotropia deviation at near(33 cm)fixation was 33.14△±8.89△ and 36.90△±10.76△, respectively(t=-2.377, P=0.019).CONCLUSIONS: Monocular occlusion test has a great impact on the exotropia deviation at near(33 cm)fixation in children with basic type of intermittent exotropia before surgery. It can expose the maximum exotropia deviation, reduce the rate of undercorrection after strabismus surgery, and provide a reliable surgical plan.

Published
2024
Full Text
View/download PDF

16. Clinical and genetic characteristics of CEL-MODY (MODY8): a literature review and screening in Chinese individuals diagnosed with early-onset type 2 diabetes

Author

Sun, Siyu, Gong, Siqian, Li, Meng, Wang, Xirui, Wang, Fang, Cai, Xiaoling, Liu, Wei, Luo, Yingying, Zhang, Simin, Zhang, Rui, Zhou, Lingli, Zhu, Yu, Ma, Yumin, Ren, Qian, Zhang, Xiuying, Chen, Jing, Chen, Ling, Wu, Jing, Gao, Leili, Zhou, Xianghai, Li, Yufeng, Zhong, Liyong, Han, Xueyao, and Ji, Linong

Published
2024
Full Text
View/download PDF

40. Artificial Intelligence for Fetal Ultrasound

Author

Cho, Hyun Cheol, Sun, Siyu, Park, Sung Wook, Kwon, Ja-Young, Seo, Jin Keun, Bock, Hans Georg, Series Editor, de Hoog, Frank, Series Editor, Friedman, Avner, Series Editor, Gupta, Arvind, Series Editor, Nachbin, André, Series Editor, Ozawa, Tohru, Series Editor, Pulleyblank, William R., Series Editor, Rusten, Torgeir, Series Editor, Santosa, Fadil, Series Editor, Seo, Jin Keun, Series Editor, and Tornberg, Anna-Karin, Series Editor

Published
2023
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results