Zheyang Li

Zheyang Li

Auto-generated profile, based on publicly available information

Zheyang Li

College of Computer Science and Technology, Zhejiang University, Hangzhou, China

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

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

Representative Publications

View All Publications
25 January 2024 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
HDHuiqi Deng
Huiqi Deng
NZNa Zou
Na Zou
MDMengnan Du
Mengnan Du
WCWeifu Chen
Weifu Chen
GFGuocan Feng
Guocan Feng
QZQuanshi Zhang
Quanshi Zhang
Various attribution methods have been developed to explain deep neural networks (DNNs) by inferring the attribution/importance/contribution score of each input variable to the final output. However, existing attribution methods are often built upon different heuristics. There remains a lack of a unified theoretical understanding of why these methods are effective and how they are related. Furthermore, there is still no universally accepted criterion to compare whether one attribution method is preferable over another. In this paper, we resort to Taylor interactions and for the first time, we discover that fourteen existing attribution methods, which define attributions based on fully different heuristics, actually share the same core mechanism. Specifically, we prove that attribution scores of input variables estimated by the fourteen attribution methods can all be mathematically reformulated as a weighted allocation of two typical types of effects, i.e. , independent effects of each input variable and interaction effects between input variables. The essential difference among these attribution methods lies in the weights of allocating different effects. Inspired by these insights, we propose three principles for fairly allocating the effects, which serve as new criteria to evaluate the faithfulness of attribution methods. In summary, this study can be considered as a new unified perspective to revisit fourteen attribution methods, which theoretically clarifies essential similarities and differences among these methods. Besides, the proposed new principles enable people to make a direct and fair comparison among different methods under the unified perspective.

22 June 2024 in Knowledge-Based Systems
Verified
Journal Article
Closed access
ZLZheyang Li
Zheyang Li
KZKai Zhang
Kai Zhang
QYQiming Yang
Qiming Yang
CLChaoxiang Lan
Chaoxiang Lan
HZHuanlong Zhang
Huanlong Zhang
SPShiliang Pu
Shiliang Pu

8 March 2023 in Applied Intelligence
Verified
Journal Article
Closed access
PDF file
XHXinyi Hu
Xinyi Hu
HFHangxiang Fang
Hangxiang Fang
LZLing Zhang
Ling Zhang
XZXue Zhang
Xue Zhang
HYHoward Hao Yang
Howard Hao Yang
Hao HuHao Hu
Hao Hu

13 September 2024 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
SCShuya Chen
Shuya Chen
ZLZheyang Li
Zheyang Li
HZHao Zhu
Hao Zhu
WTWenming Tan
Wenming Tan
YRYe Ren
Ye Ren
ZXZhiyu Xiang
Zhiyu Xiang
The performance of neural radiance field technologies deteriorates rapidly when sparse views are used as input. In this paper, we propose a simulated viewpoint enhancement for surface reconstruction that extracts diverse geometric features from the depth to address this limitation. We design a novel pseudo-view generation approach to simulate dense viewpoints, in which the RGB-D pairs are converted into new images corresponding to the translational poses through reprojection. To address potential holes in the pseudo-views resulting from occlusion in the reprojecting process, we propose a hole handling mechanism to exclude them during training. Additionally, normal information is derived from the depth, serving as supervision to enhance the scene geometry. Experiments conducted on various baseline models and datasets demonstrate that our algorithm effectively utilizes the inherent geometry embedded in the depth maps, leading to a significant improvement in the quality of 3D surface reconstruction compared to existing methods.

12 March 2024 in arXiv
Verified
Preprint
Open access
PDF file
LYLongrong Yang
Longrong Yang
XZXianpan Zhou
Xianpan Zhou
XLXuewei Li
Xuewei Li
LQLiang Qiao
Liang Qiao
ZLZheyang Li
Zheyang Li
XLXi Li
Xi Li
Knowledge distillation (KD) has shown potential for learning compact models in dense object detection. However, the commonly used softmax-based distillation ignores the absolute classification scores for individual categories. Thus, the optimum of the distillation loss does not necessarily lead to the optimal student classification scores for dense object detectors. This cross-task protocol inconsistency is critical, especially for dense object detectors, since the foreground categories are extremely imbalanced. To address the issue of protocol differences between distillation and classification, we propose a novel distillation method with cross-task consistent protocols, tailored for the dense object detection. For classification distillation, we address the cross-task protocol inconsistency problem by formulating the classification logit maps in both teacher and student models as multiple binary-classification maps and applying a binary-classification distillation loss to each map. For localization distillation, we design an IoU-based Localization Distillation Loss that is free from specific network structures and can be compared with existing localization distillation losses. Our proposed method is simple but effective, and experimental results demonstrate its superiority over existing methods. Code is available at https://github.com/TinyTigerPan/BCKD.

6 December 2025 in Knowledge-Based Systems
Verified
Journal Article
Closed access
KZKai Zhang
Kai Zhang
CLChaoxiang Lan
Chaoxiang Lan
YXYazhang Xu
Yazhang Xu
ZLZheyang Li
Zheyang Li
WTWenming Tan
Wenming Tan
Jilin HuJilin Hu
Jilin Hu