1 code implementation • 2 Apr 2024 • Biao Jiang, Xin Chen, Chi Zhang, Fukun Yin, Zhuoyuan Li, Gang Yu, Jiayuan Fan
However, this proficiency remains largely unexplored in other multimodal generative models, particularly in human motion models.
no code implementations • 18 Mar 2024 • Zhuoyuan Li, Zikun Yuan, Li Li, Dong Liu, Xiaohu Tang, Feng Wu
Moreover, segmentation mask is considered in the joint rate-distortion optimization for motion estimation and partition estimation to derive the motion vector of different regions and partition more accurately.
1 code implementation • 18 Sep 2023 • Zhuoyuan Li, Bin Dong, Pingwen Zhang
Data assimilation is crucial in a wide range of applications, but it often faces challenges such as high computational costs due to data dimensionality and incomplete understanding of underlying mechanisms.
no code implementations • 26 Jul 2023 • Xiang Huang, Zhuoyuan Li, Hongsheng Liu, Zidong Wang, Hongye Zhou, Bin Dong, Bei Hua
Recently, using neural networks to simulate spatio-temporal dynamics has received a lot of attention.
no code implementations • 11 Jul 2023 • Chuanbo Tang, Xihua Sheng, Zhuoyuan Li, Haotian Zhang, Li Li, Dong Liu
In the offline stage, we fine-tune a trained optical flow estimation network with the motion information provided by a traditional (non-deep) video compression scheme, e. g. H. 266/VVC, as we believe the motion information of H. 266/VVC achieves a better rate-distortion trade-off.
no code implementations • 24 Feb 2022 • Kelei He, Chen Gan, Zhuoyuan Li, Islem Rekik, Zihao Yin, Wen Ji, Yang Gao, Qian Wang, Junfeng Zhang, Dinggang Shen
Transformers have dominated the field of natural language processing, and recently impacted the computer vision area.
1 code implementation • 17 May 2021 • Kelei He, Wen Ji, Tao Zhou, Zhuoyuan Li, Jing Huo, Xin Zhang, Yang Gao, Dinggang Shen, Bing Zhang, Junfeng Zhang
Specifically, a bidirectional image synthesis and segmentation module is proposed to segment the brain tumor using the intermediate data distributions generated for the two domains, which includes an image-to-image translator and a shared-weighted segmentation network.