爆走黑料

This project explores the design and implementation of Genetic Programming (GP) for facial expression regression with an emphasis on hardware acceleration using Field-Programmable Gate Array (FPGA). Regression tasks such as age estimation or head pose prediction from facial images are typically addressed with deep learning. However, deep models are computationally expensive and often lack interpretability. GP offers an interpretable symbolic alternative, and FPGA hardware provides energy-efficient acceleration for the large number of evaluations required during evolution. This project will pursue the following three objectives: develop a GP framework for facial image regression tasks; implement key GP components on FPGA (fitness evaluation, tree evaluation) to accelerate training; compare FPGA-based GP with CPU-only GP in terms of execution speed, energy efficiency, and scalability.