Discover companies you will love
Hello, I'm Wang Fengyusheng. I'm sorry but I have just learned Japanese for half a year. こにちは、私は王風宇声です。すみませんが、私は日本語を学んで半年になるところです。
I like to face challenges in Computer Vision and VR/AR. 私はコンピュータビジョンやVR / ARに関する挑戦に直面することが好きです。
Conducted a preliminary study on improving Phase Detection Autofocus using neural networks: · Captured a dataset of 3000+ raw images across 20+ scenarios using one lens, and preprocessed an additional dataset of 4000+ raw images across 7 scenarios obtained from a vendor us
1. Explored the possibility of low-cost real-time 3D SLAM based on RGBD sensors: · Utilized Kinect V1 and GTX 980Ti (working remotely during COVID with these spared devices provided by supervisor).
Ported Octree-based CNN (O-CNN) and Adaptive O-CNN from Caffe to TensorFlow and improved it: · Enabled a resolution of 63356^3, up from the original 256^3. · Ported all the operators on octrees and adaptive octrees including convo
1. Participated in C++ development of the adaption of MXNet framework for Cambricon’s machine learning processors. 2. Tested the inference of SSD and Faster-RCNN networks under several modes (single-core and multi-core, dense parameters and sparse parameters) when running Python scripts and binary executables compiled from C++.