Chuangchuang Zhou

Master Thesis: Multimodal learning for physical properties of containers: - Collected a large-scale multimodal dataset with a focus on the perception task for robotic pouring, augmented the audio data - Evaluated the performances of different network structures, implemented multimodal pouring networks based on audition and haptics to predict liquid height robustly in containers in noisy environments - Verified and compared the reliability and robustness of pouring networks across various robotic experiments, generated the shape of unknown containers by combining multimodalities - Segmented and tracked cup in video streaming for dynamics human-to-robot handovers Programming language and tools: Python, PyTorch, ROS, OpenCV

- Preprocessed the images, such as detected vehicle screen display. - Designed and implemented algorithms for camera calibration and eliminated the effects of distortion - Built 3D model by OpenGL, compared its projection and original image to verify the influences of rear-view camera’s positions - Implemented a GUI to make operation more intuitive for different vehicles Programming language and tools: Python, OpenCV, PyQt5, OpenGL, Blender

- Analyzed and compared the algorithms of point cloud registration and recognition - Optimized a global point cloud registration algorithm to reduce the registration time, applied it with ICP to improve the accuracy of recognition - Measured different workpieces with industrial tomography, analyzed and visualized industrial computed tomography data by using VGSTUDIO MAX Programming language and tools: C++, CMake, PCL