Yuxin Yao

Yuxin Yao

PhD student in Engineering Department, University of Cambridge

Yuxin Yao

Welcome to my website. I am Yuxin Yao, a 2nd Year PhD student in Information Engineering at University of Cambridge, supervised by Prof. Joan Lasenby. I obtained my Master of Engineering degree integrated with my bachelor at University College London.

My current research focus on computer vision. My research interests includes:

  1. 4D Reconstruction and Novel View Synthesis with Gaussian Splatting.
  2. Geometric Deep Learning.
  3. Generative AI.

I am currently seeking for collaborations on project in these fields. Please contact me if you are interested!

Project Experience:

Simplifying and Generalising Equivariant Geometric Algebra Networks

Supervised By: Professor Joan Lasenby at Cambridge, collaborated with Christian Hockey

  • Developed a simplified and generalized equivariant Geometric Algebra Transformer (CGATr) with a generalized signature by removing constraints on geometric signatures.
  • Analyzed and experimented with the Geometric Algebra Transformer (GATr) to study the effectiveness of its components.
  • Applied CGATr to protein structure prediction, N-body dynamics, and camera pose estimation, showcasing its potential in geometric deep learning.
  • Presented this work at the 9th Conference on Applied Geometric Algebras in Computer Science and Engineering (Amsterdam, NL).
  • Protein Structure Prediction Comparison between CGATr and common neural network: Predicted protein structure

Unsupervised Visual Relocalization

Supervised By: Professor Simon Julier at UCL as final year project

  • Implemented the method in Unsupervised Metric Relocalization Using Transform Consistency Loss
  • Utilized direct image alignment to find the relative transformation between the query image and reference images. Employed the Gauss-Newton method in searching optimized transformations between the feature maps of the query and reference image.
  • Generated dataset with CARLA for training/testing.
  • Constructed U-Net inferring feature maps and saliency maps of 2D images, extracting important features and masking out moving objects occluding the features.
  • Image and Saliency map example, the saliency map masks the fast moving part: Feature map and Saliency map example:

Human Motion Prediction on Egocentric Dataset

Supervised By: Professor Siyu Tang at ETH Zurich Computer Vision and Learning Group

  • Trained a motion prior of the egocentric dataset EgoBody which recorded the first person perspective video with Holo-lens under social interaction scenario. Predicted the motion for future 8 or 9 frames given the body model of the first 1 or 2 frames.
  • Familarized and used smpl-x and smpl model. Applied AMASS and Egobody dataset to GAMMA model for a body regressor and marker predictor, which uses conditional VAE with DLow method, and GRU.

For more previous projects and detailed description, please check my CV

Publication

  • Hockey, C., Yao, Y., Lasenby, J. Simplifying and Generalising Equivariant Geometric Algebra Networks. The 9th conference on Applied Geometric Algebras in Computer Science and Engineering (Abstract accepted )

  • Chen, H., Li, Z., Yao, Y. (2022, November). Multi-agent reinforcement learning for fleet management: a survey. In 2nd International Conference on Artificial Intelligence, Automation, and High-Performance Computing (AIAHPC 2022) (Vol. 12348, pp. 611-624). SPIE.

  • Yan, Y., Schaffter, T., Bergquist, T., …Yao, Y..,… DREAM Challenge Consortium. (2021). A Continuously Benchmarked and Crowdsourced Challenge for Rapid Development and Evaluation of Models to Predict COVID-19 Diagnosis and Hospitalization. JAMA network open, 4(10), e2124946-e2124946.