Yuncheng Li

Yuncheng Li

Auto-generated profile, based on publicly available information

Yuncheng Li

University of Rochester

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 March 2017 in IEEE Transactions on Multimedia
Verified
Journal Article
Closed access
YLYuncheng Li
Yuncheng Li
LCLiangliang Cao
Liangliang Cao
JZJiang Zhu
Jiang Zhu
JLJiebo Luo
Jiebo Luo
Composing fashion outfits involves deep under-standing of fashion standards while incorporating creativity for choosing multiple fashion items (e.g., jewelry, bag, pants, dress). In fashion websites, popular or high-quality fashion outfits are usually designed by fashion experts and followed by large audiences. In this paper, we propose a machine learning system to compose fashion outfits automatically. The core of the proposed automatic composition system is to score fashion outfit candidates based on the appearances and metadata. We propose to leverage outfit popularity on fashion-oriented websites to supervise the scoring component. The scoring component is a multimodal multiinstance deep learning system that evaluates instance aesthetics and set compatibility simultaneously. In order to train and evaluate the proposed composition system, we have collected a large-scale fashion outfit dataset with 195K outfits and 368K fashion items from Polyvore. Although the fashion outfit scoring and composition is rather challenging, we have achieved an AUC of 85% for the scoring component, and an accuracy of 77% for a constrained composition task.

9 December 2013
Verified
Conference Paper
Open access
PDF file
YSYoung Chol Song
Young Chol Song
HKHenry Kautz
Henry Kautz
JAJames Allen
James Allen
MSMary Swift
Mary Swift
YLYuncheng Li
Yuncheng Li
Ce ZhangCe Zhang
Ce Zhang

1 June 2017 in arXiv
Verified
Preprint
Open access
PDF file
YLYuncheng Li
Yuncheng Li
YSYale Song
Yale Song
JLJiebo Luo
Jiebo Luo
Learning to rank has recently emerged as an attractive technique to train deep convolutional neural networks for various computer vision tasks. Pairwise ranking, in particular, has been successful in multi-label image classification, achieving state-of-the-art results on various benchmarks. However, most existing approaches use the hinge loss to train their models, which is non-smooth and thus is difficult to optimize especially with deep networks. Furthermore, they employ simple heuristics, such as top-k or thresholding, to determine which labels to include in the output from a ranked list of labels, which limits their use in the real-world setting. In this work, we propose two techniques to improve pairwise ranking based multi-label image classification: (1) we propose a novel loss function for pairwise ranking, which is smooth everywhere and thus is easier to optimize; and (2) we incorporate a label decision module into the model, estimating the optimal confidence thresholds for each visual concept. We provide theoretical analyses of our loss function in the Bayes consistency and risk minimization framework, and show its benefit over existing pairwise ranking formulations. We demonstrate the effectiveness of our approach on three large-scale datasets, VOC2007, NUS-WIDE and MS-COCO, achieving the best reported results in the literature.

10 March 2016 in arXiv
Verified
Preprint
Open access
PDF file
YWYu Wang
Yu Wang
YLYuncheng Li
Yuncheng Li
JLJiebo Luo
Jiebo Luo
In this paper, we study follower demographics of Donald Trump and Hillary Clinton, the two leading candidates in the 2016 U.S. presidential race. We build a unique dataset US2016, which includes the number of followers for each candidate from September 17, 2015 to December 22, 2015. US2016 also includes the geographical location of these followers, the number of their own followers and, very importantly, the profile image of each follower. We use individuals' number of followers and profile images to analyze four dimensions of follower demographics: social status, gender, race and age. Our study shows that in terms of social influence, the Trumpists are more polarized than the Clintonists: they tend to have either a lot of influence or little influence. We also find that compared with the Clintonists, the Trumpists are more likely to be either very young or very old. Our study finds no gender affinity effect for Clinton in the Twitter sphere, but we do find that the Clintonists are more racially diverse.

10 March 2016 in arXiv
Verified
Preprint
Open access
PDF file
YWYu Wang
Yu Wang
JLJiebo Luo
Jiebo Luo
RNRichard Niemi
Richard Niemi
YLYuncheng Li
Yuncheng Li
THTianran Hu
Tianran Hu
In this paper, we propose a framework to infer the topic preferences of Donald Trump's followers on Twitter. We first use latent Dirichlet allocation (LDA) to derive the weighted mixture of topics for each Trump tweet. Then we use negative binomial regression to model the "likes," with the weights of each topic serving as explanatory variables. Our study shows that attacking Democrats such as President Obama and former Secretary of State Hillary Clinton earns Trump the most "likes." Our framework of inference is generalizable to the study of other politicians.

18 September 2015 in arXiv
Verified
Preprint
Open access
PDF file
YLYuncheng Li
Yuncheng Li
YCYang Cong
Yang Cong
TMTao Mei
Tao Mei
JLJiebo Luo
Jiebo Luo
Most state-of-the-art image retrieval and recommendation systems predominantly focus on individual images. In contrast, socially curated image collections, condensing distinctive yet coherent images into one set, are largely overlooked by the research communities. In this paper, we aim to design a novel recommendation system that can provide users with image collections relevant to individual personal preferences and interests. To this end, two key issues need to be addressed, i.e., image collection modeling and similarity measurement. For image collection modeling, we consider each image collection as a whole in a group sparse reconstruction framework and extract concise collection descriptors given the pretrained dictionaries. We then consider image collection recommendation as a dynamic similarity measurement problem in response to user's clicked image set, and employ a metric learner to measure the similarity between the image collection and the clicked image set. As there is no previous work directly comparable to this study, we implement several competitive baselines and related methods for comparison. The evaluations on a large scale Pinterest data set have validated the effectiveness of our proposed methods for modeling and recommending image collections.