Biography
My name is Zixing Lei(雷梓行) and you can just call me Elwood. Currently I am a master student in the Cooperative Medianet Innovation Center(CMIC) of Shanghai Jiao Tong University(SJTU), supervised by Prof. Siheng Chen. My research interests include computer vision, embodied AI and multi-robots perception.
Interests
- Computer Vision
- Embodied AI
- Multi-Modality 3D Understanding
Education
-
MEng in Electronic Information, 2021-2024
Shanghai Jiao Tong Univerisity
-
BSc in Information Engineering, 2017-2021
Shanghai Jiao Tong Univerisity
Recent Publications
Quickly discover relevant content by filtering publications.
![[ICRA 2024] Robust Collaborative Perception without External Localization and Clock Devices](/publication/conference-paper/freealign/featured_hu9e2cad90886f3ef9104988c22f7750f7_277264_150x0_resize_q75_h2_lanczos_3.webp){kind=small}
![[NeurIPS2023] Emergent Communication in Interactive Sketch Question Answering](/publication/conference-paper/ecisqa/featured_hu2ef339ea6667f10a43937fefae448226_168158_150x0_resize_q75_h2_lanczos_3.webp){kind=small}
![[NeurIPS2022] Where2comm: Communication-Efficient Collaborative Perception via Spatial Confidence Maps](/publication/conference-paper/where2comm/featured_hu37bc554eeab7d72d04b7620a327a2899_1535997_150x0_resize_q75_h2_lanczos_3.webp)
![[ECCV2022] Latency-Aware Collaborative Perception](/publication/conference-paper/syncnet/featured_huedc3d39f26bb99a11cdaff5fff7dd3c4_854291_150x0_resize_q75_h2_lanczos_3.webp)
Experience
Graphic Software Engineering Intern
- Developed an automatic tools for Intel DG2 GPU to test the quality of encoded gaming video and adaptively adjust encoding parameters for Intel cloud gaming and live-streaming experience.
- Developed VMAF for gaming, a machine learning model to evaluate encoded video by learning human audience judgement specifically tuning on videos of popular gaming such as league of legends and overwatch.
Video ReID of Endangered Amur Tigers
- Collect video data in order to ReID amur tigers with deep learning methods to help to track them in natural reserve.
- Established an amur tiger ReID dataset with object detection and Multiple Object Tracking methods.
- Created a video ReID method which is 7% more accurate than the previous image ReID method for amur tigers.
MLLM Engineer
- I am leveraging the excited vision language model to improve the performance of the governance system in TikTok.
- I am wokring on the efficient, robust and interpretable large multi-modal models, especially focus on COT technology.
