I am a PhD student at INSAIT in Sofia, Bulgaria, working under the supervision of Dr. Danda Paudel and Prof. Luc Van Gool. My research focuses on 3D scene understanding and generation.
Previously, I earned my MRes in AI and Machine Learning at Imperial College London, where I explored camera relocalization, and geometry-based machine learning under the guidance of Dr. Tolga Birdal. I also hold a B.Sc. in Automation from Beijing Institute of Technology. Additionally, I am an active member of AnySyn3D, a research community towards 3D.
Previously, I gained experience working in both academia and industry, including research roles at Pixomondo/Sony as a research scientist and a computer vision research intern at DISCOVER Lab, AIR, Tsinghua University, supervised by Dr. Zhao Hao.
Apart from my research, I am a swimming fancier, especially in crawl and backstroke, a founder and main member of BIT swimming association and BIT Swimming Team during 2020 to 2022. Everyone is welcomed to join us!
✨News🔥
10/2024 🎉 I am happy to announce that I graduated from Imperial College London and will join INSAIT to continue as a PhD student.
06/2024 🎧 I will be working at a part-time research scientist at Innovation Lab, Pixomondo, Sony, concentrating on the neural radiance fields and Gaussian splatting for VFX industry.
09/2023 🥳 I have enrolled in Imperial College London as Postgraduate Research Students (MRes in AI and Machine Learning) and I am actively looking for PhD positions.
07/2023 🔥 Our paper MARS has been accepted by CICAI 2023 and won best paper runner-up award. See project page here.
Large-scale Gaussian splatting based dataset for scene understanding. We propose a new dataset with 3D Gaussian splatting and a vision-language pretraining framework to learn the scene understanding task.
Project homepage / arxiv / Code
Context-aware pose distributional localization. Given an ambiguous text description, our method accurately estimates the camera pose distribution across a large-scale urban environment.
Project homepage / arxiv / Code
GaussianGrasper is a robotic system that uses 3D Gaussian Splatting and an Efficient Feature Distillation module to enable robots to grasp objects based on language instructions.
Project homepage / arxiv / Code
We propose A newly proposed primitive pruning framework for Gaussian fields based upon the spectrum of primitive graphs; And A novel feature splatting and mixing module to compensate for the performance drop caused by the pruning; Reached state-of-the-art results, in terms of both quality and speed, on various benchmarks with low memory footprint.
Project homepage / arxiv / Code
• Outstanding Graduate of Beijing Institute of Technology
• Outstanding Individual of Beijing Institute of Technology
• Student Representation of School of Automation, Beijing Institute of Technology
• Student Representation of sports clubs at Beijing Institute of Technology
• CASC Scholarship, China Aerospace Science and Technology Corporation, Oct 2021
• Outstanding Individual for 2020-2021 Academic Year, BIT, Oct 2021
• First Prize, BIT Balance Car Competition, Jun 2021
• Academic Excellence Scholarship x 8 (Top 10%)
• Best Design Award, 6th Smart Car Competition
• Outstanding Graduate of Beijing Institute of Technology
• Outstanding Individual of Beijing Institute of Technology
• Student Representation of School of Automation, Beijing Institute of Technology
• Student Representation of sports clubs at Beijing Institute of Technology
• CASC Scholarship, China Aerospace Science and Technology Corporation, Oct 2021
• Outstanding Individual for 2020-2021 Academic Year, BIT, Oct 2021
• First Prize, BIT Balance Car Competition, Jun 2021
• Academic Excellence Scholarship x 8 (Top 10%)
• Best Design Award, 6th Smart Car Competition