Intern, University of North Carolina at Charlotte (UNCC),
– Designing and verifying a signal enhancement and physical augmentation framework based on Reinforcement Learning, which will not only enhance the generalization of the model and solve many unseen activities of the dataset, but also add real physical constraints of the human body through Reinforcement Learning methods.
Intern, University of North Carolina at Charlotte (UNCC),
– Design a plug-and-play physics-based optimization module according to the physical characteristics of human movement.
– Module can refined the point cloud data collected by millimeter-wave radar based on the characteristics of human movement to meet the physical constraints
Lab Research, Nanjing University of Aeronautics and Astronautics, department of Automation, China, Nanjing
– Design a novel sparse log-step gradient prior which adopts a mixture of logarithm and step functions to regularize the image gradients and combine it with the Poisson distribution to formulate the blind multi-image deconvolution problem.
– Incorporate the methods of variable splitting and Lagrange multiplier to convert the original problem into sub-problems.
– Design a non-blind multi-image deconvolution algorithm which is based on the log-step gradient prior to reach the final restored image.
Lab Project, Nanjing University of Aeronautics and Astronautics, department of Automation, China, Nanjing
– Organized merchandise from numerous clients preparing them to be shot in photo studio.
– Build an end-to-end dehazing network based on the SwinTransformer model under the GAN architecture, and conduct training and verification on the dehazing dataset.
– Reduce the amount of model parameters and calculations through cheap convolution, network clipping, model quantification and other model lightweight technologies.
– Combine the dehazing network with the YOLOv5 target The target detection network is deployed on NVIDIA edge computing devices.