UROP Proceedings 2020-21

School of Engineering Department of Electronic and Computer Engineering 154 Artificial Intelligence Methods for Medical Videos Supervisor: LI Xiaomeng / ECE Student: TSANG King Hoi / COMP Course: UROP1100, Summer This report mainly reviews the basic concepts of deep learning and various network architectures used in the context image and video analysis that was explored throughout the course of this project. Some of the concepts are then used to design a network architecture to perform classification and localization tasks and evaluated using the ESAD and MESAD dataset. Concepts discussed includes convolutional neural networks, residual neural networks, the YOLO algorithm residual neural networks and Feature pyramids. These concepts will be used to explain different architectures used for image and video analysis explored throughout the duration of this project, such as the RetinaNet network and examples of semi-supervised networks for segmentation tasks. End-to-End Instance Segmentation with Transformers Supervisor: LI Xiaomeng / ECE Student: YU Yiduo / COMP XIANG Tianqi / COSC Course: UROP1000, Summer UROP1100, Summer This is a report of professor LI Xiaomeng's UROP project End-to-End Instance Segmentation with Transformers in Summer 2021. In this project, we are instructed and guided to dig into both the novel technologies of instance segmentation and the application of those networks. Some fundamental classical papers which laid a solid base for the image segmentation task are reviewed. This report briefly introduces the general outlines of our work during the summer. Graph Neural Network for Drug Discovery Supervisor: LI Xiaomeng / ECE Student: YANG Zheng / ELEC Course: UROP1100, Spring After a brief introduction of this project, it goes to the main part of this report. First for the related work parts I talk about my own understandings of the CNN and Transformer architecture after a semester’s study. Then I introduce two papers about combination of CNN and Transformer on medical images process. For this part, I mainly talk about their architecture respectively and compare their difference. The mainly part ends with the conclusions which is my feelings about this field after three months’ study. The last part is the reflection about this whole semester’s study including the problems I have.

RkJQdWJsaXNoZXIy NDk5Njg=