Li Jin

金立

Ph.D. student
School of Software
Shandong University
Email: jinli@mail.sdu.edu.cn

I am a third-year PhD student at Shandong University under the supervision of Prof. Xueying Qin. I currently intern at X Square Robot. My past experiences include working as a visiting PhD researcher at the Visual Computing and Learning Lab (VCL), Peking University (led by Prof. Baoquan Chen), and completing an internship at Tencent.

Object pose estimation for 3D perception (instance-level & category-level)
Open-world 3D object perception (shape, pose, semantics, material)
Construction of multi-attribute large-scale datasets for open-world 3D perception models

I am currently seeking a full-time position.

Publications

*: equal contribution; † corresponding author(s)

	CoSMo3D: Open-World Promptable 3D Semantic Part Segmentation through LLM-Guided Canonical Spatial Modeling Li Jin, Weikai Chen, Yujie Wang, Yingda Yin, Zeyu Hu, Runze Zhang, Keyang Luo, Shengju Qian, Xin Wang, Xueying Qin CVPR 2026 (Oral Presentation, Award Candidate, Top 1.81%) project page / code We propose CoSMo3D for open-world promptable 3D semantic part segmentation. By constructing a unified canonical dataset via LLM-guided alignment and inducing a latent canonical reference frame through a dual-branch architecture, CoSMo3D achieves state-of-the-art performance with more stable and transferable part semantics across poses and categories.
	One-shot 3D Object Canonicalization based on Geometric and Semantic Consistency Li Jin, Yujie Wang, Wenzheng Chen, Qiyu Dai, Qingzhe Gao, Xueying Qin, Baoquan Chen^† CVPR 2025 (Highlight Presentation, Top 13.5%) project page / code We introduce a one-shot category-level 3D object canonicalization method that synergistically combines geometric priors and semantic cues extracted from large language models (LLMs) and vision-language models (VLMs). Additionally, we construct the Canonical Objaverse Dataset (COD), which contains 32,000 samples.
	Prior-free 3D Object Tracking Xiuqiang Song, Li Jin, Zhang Zhengxian, Jiachen Li, Fan Zhong, Guofeng Zhang, Xueying Qin^† CVPR 2025 (Highlight Presentation, Top 13.5%) code We propose a novel prior-free 3D object tracking method that operates without relying on any models or training data. This method consists of a geometry generation module and a pose optimization module, which can automatically and iteratively enhance each other to gradually construct the necessary information for tracking.
	A Visual Servo System for Robotic On-Orbit Servicing Based on 3D Perception of Non-Cooperative Satellite Panpan Zhao, Li Jin, Yeheng Chen, Jiachen Li, Xiuqiang Song, Wenxuan Chen, Nan Li, Wenjuan Du, Ke Ma, Xiaokun Wang, Yuehua Li, Xiangxu Meng, Xueying Qin^† ICRA 2025 We present a 3D perception-based visual servo system for non-cooperative satellites. This system integrates reconstruction and tracking, using an alternating iterative strategy to enhance shape perception and pose estimation accuracy in complex orbital conditions.
	CSS-Net: Domain Generalization in Category-level Pose Estimation via Corresponding Structural Superpoints Li Jin, Xibin Song, Jia Li, Changhe Tu, Xueying Qin^† ICME 2024 (Oral Presentation, Top 15%) We propose a domain generalization method for category-level pose estimation based on structural superpoints. Trained solely on simulated data, this method effectively generalizes to unseen domain distributions in real datasets.
	Implicit Coarse-to-Fine 3D Perception for Category-level Object Pose Estimation from Monocular RGB Image Jia Li, Li Jin, Xibin Song, Yeheng Chen, Nan Li, Xueying Qin^† ICRA 2024 We propose a category-level pose estimation framework from a single RGB image—Feature Auxiliary Perception Network (FAP-Net). This method aims to address the ambiguity in object translation and size caused by relying solely on RGB images.
	Online Hand-Eye Calibration with Decoupling by 3D Textureless Object Tracking Li Jin, Kang Xie, Wenxuan Chen, Xin Cao, Yuehua Li, Jiachen Li, Jiankai Qian, Xueying Qin^† ICRA 2023 We propose a novel hand-eye calibration method based on the natural 3D object, which can work online and automatically even if the object is textureless or weakly textured.

Website template courtesy of Jon Barron.