Xingfu du

Background

About Xingfu du

As a software engineer at SenseTime, a leading artificial intelligence company, I have been developing innovative solutions for various domains, such as computer vision, neural network, and computational photography, for over a year. I am passionate about applying my strong theoretical background, data analysis skills, and creative problem-solving skills to create cutting-edge products that enhance the user experience and deliver value to the market. I am currently part of the AIPE Nightsight fusion team, where I participate in delivering a software development kit that implements multi-frame noise reduction (MFNR) and HDR fusion in low-light conditions. I have implemented algorithm modules that significantly increase SNR and reduce motion blur, as well as developed CPU and GPU codes that greatly improve algorithm efficiency. In AIGC team since the beginning of 2023, I've implemented the stable diffusion model with qualcomm neutral processing SDK, which is a state-of-art text-to-image model developed by Qualcomm AI reaserch team that has the world's first on-device demonstration on Android phone. We have optimized and reduced the runtime of the whole pipeline down to around 15s. I have also contributed to refactoring and maintaining RGBW, RGBCMY, and RGBW Fusion code bases, as well as building documentation and testing tools. Previously, I was involved in delivering a fixed-point C++ reference model for RGBCMY full-resolution pipeline project, which implemented RAISR, a rapid and efficient learning-based method of generating super-resolution images. I have completed various modules, conducted modular tests, and implemented multi-threading in each module.