在此处输入标题

Josie

As a senior undergraduate student in computer science school, XiDian University, China, I am passionate about computer vision and deep learning, especially the application of deep learning in Objection Detection, Segmentation and Recognition. Additionally I am doing some work on Generative Adversarial Networks. Recently I noticed your lab's work on 3D reconstruction by SfM. I hope that I could have the opportunity to work with Professor Crandall on the research of Object Recognition and Generative Adversarial Networks as an intern in the IUB Computer Vision Lab next summer.

Mathematics, the foundation of computer science, appeals to me because of its abstract beauty. My collaboration and learning skills have been further enhanced through mathematical contest in modeling. And I got Meritorious Winner in MCM/ICM(Mathematical Contest in Modeling) 2018. My teammates and I modeled a Propagation Path-Loss Model. This model focuses on the analysis of the propagation path and transmission loss of HF sky wave, including the propagation loss in free space and ionosphere and reflection loss over the sea-surface. After the competition, my teammates and I summarized the solutions in this paper, 'The Design and Implement of Propagation Path-Loss Model over Sea-Surface under HF'. Our paper is here: http://www.dpi-proceedings.com/index.php/dtcse/article/view/24724

I want to do more research in Computer Vision. Recently I studied object semantic analysis and I'm working in Object Detection and Segmentation of road images as part of the 3D reconstruction of traffic scenes. In order to build the model, which is designed to detect and segment the vehicle, pedestrian and other objects in the foreground of road images, I read papers about R-CNN, SPP-Net, Fast R-CNN, Faster R-CNN, Mask R-CNN, YOLO and polygon RNN++, and some of them has been summarised in my blog. These ingenious designs, such as ResNet, RPN, ROI Align and polygon segmentation, have achieved very accurate results in Object Detection and Segmentation, and they will be applied to my model. Furthermore, maybe I can work on Object Detection and Segmentation as part of Crandall's work on Object Recognition.

In addition to Object Detection and Segmentation, I have also been working on Generative Adversarial Networks. A summary and some ideas of GANs can be found in my blog. I am trying to use GAN to generate traffic scenes of different weather and seasons, which is a part of the reconstructed traffic scenes model. I have improved Cycle GAN to generate the rectangular winter road images. It will also be used in experiments on transition of other seasons. And my progress will be updated on GitHub. I read recent research on GANs too, like DCGAN, Self-Attention GAN, big GAN. The pictures, music and poetry they generated are amazing, but I think we can still make some improvements.

The notes and processes of my work can be found in my GitHub and blog:
Github profile: https://github.com/JosieHong
Blog: https://josiehong.github.io/

Based on the above knowledge and experience, I think I am qualified to become a research assistant in the field of computer vision to do more constructive work. I have a passion for promoting the development of computer vision and letting the computer see and understand the world better. This internship will be a great start for all of this.