Fangzheng Tian is a PhD of Department of Computer Science & Artificial Intelligence at Jeonbuk National University. His research interests include machine learning, computer vision, digital human, SLAM, depth estimation, 3D reconstruction and medical image processing.
Email: tian_fangzheng@jbnu.ac.kr  /  Github  /  CSDN  / 
Using UAV instead of manual work, based on multi-sensor, the UAV has the function of autonomous positioning and navigation in the room to collect data, and further combines machine vision with deep learning to realize defect detection on the surface of the aircraft, which is mainly responsible for the autonomous positioning of the UAV.

Use the camera at the end of the arm to guide the arm to tap the keyboard precisely and output the corresponding characters. The project won the 3rd prize of "Big Airplane Artificial Intelligence Innovation Application Competition"(project leader)

The CT images of patients were sent into the pre-trained machine learning model after a series of processing to give information such as whether the physical examination subject suffered from gastric cancer, accurately distinguish the pathological stage of gastric cancer and so on, and achieve T-stage diagnosis of gastric cancer. The project won the 2nd prize of the “The China Graduate Electronics Design Contest” and the 3rd prize of the “Smart City Competition”

Use the detection algorithm to detect irregular behaviors in the kitchen in real time, such as smoking, not wearing a mask, not wearing a hat, rats, etc. The algorithm is installed on the TX2 device.

Mosaic is regarded as a multi image matching problem, and the invariant local features are used to find the matching between all images. Our method is insensitive to the order, direction, proportion and illumination of the input image, and can recognize multiple panoramas in the disordered image dataset.

Mainly responsible for the compilation of target detection algorithms, including Faster R-CNN, YOLO, M2Det and license plate recognition cases.

Perform image segmentation, three-dimensional reconstruction, and blood vessel center line extraction on the CT images in the physical examination, and diagnose whether the examination has coronary heart disease by calculating the value of FFRCT.

Based on the personal photo database imported in advance, the camera is used to obtain the video, and the best match of the face is achieved based on rapid detection of the face to achieve face recognition.

By using pure image processing technology to correct the angle of the captured image, restoring the forward view shooting effect.