ABOUT
I enjoy to see what the deep learning can bring us the amazing experience. Hence, I focus my research in deep learning and computer vision. In addition, I love to explore the new backend skills, such as Container, Distributed System, and Database. Therefore, I want to bring my positive impact to our society.
2D/3D object detection
Image Segmentation
GAN/RL
LSTM
Computer Vision
Distributed system
Unlocking Life’s Mysteries
3D OBJECT DETECTION
July 10, 2019
Frustum PointNet is used to solve the object detection problem in 3D coordinates.
The Lidar cloud points are projected from 3D to 2D camera coordinates. Using the Frustum to restrict these points. Based on the object detection, it uses this to locate the interested points as the training data.
After training, we can locate the objects in 3D coordinates from LiDar data with 2D data.
SPEED ESTIMATION
July 29, 2019
I joined the speed estimation challenge held by comma.ai Inc. I use the end-to-end learning model from Nvidia and Optical flow as image preprocessing.
I conduct two experience with(improve from 2.40 to 1.0) or without the Kalman Filter.
OPEN SOURCE
Apache Kafka
I made two contributions to the open source project, Kafka.
KAFKA-4772: Exploit #peek to implement #print() and other methods #2955
KAFKA-4830: Augment KStream.print() to allow users pass in extra parameters in the printed string #3085
I proposed a new feature, KIP-160, allowing users to pass extra parameters in the printed string to improve debugging, released on Kafka-1.10.