Welcome to visit netease Cloud Community to learn more about Netease’s technical product operation experience.
Internship experience of Netease Hangyan Yidun
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Internship period: December 2017 to June 2018
Intern Department: Netease Yidun Computer vision Algorithm Group
Time flies, and the six-month internship is fleeting. Take notes while the iron is hot, share a few tips, and give yourself some personality (haha).
Since last December, I have been working as an intern of deep learning-computer vision algorithm in AI Group (also known as Yidun Computer vision Algorithm Group) of Information Security Department of netease Hangzhou Research Institute (I was very excited at that time). Before I came here, I was a complete blank, looking at the cool demos of others in the group, and I was envious. At the beginning, I worked with an official employee in OCR related work and learned from target detection model. The first paper I read at that time was Faster R-CNN. I still remember that I was completely confused when I read it for the first time and couldn’t understand what I was saying at all. Later, my former tutor sent me a wechat tweet introducing the research progress of target detection based on deep learning. I had a general understanding of the target detection field first, and then started from R-CNN, spP-NET, FAST R-CNN, and then faster R-CNN. They are all series based on region proposals, which can be said to be one of the most important branches in the field of target detection. The most important idea in faster R-CNN is the anchor generation mechanism, which proposes a relatively region proposal network(RPN) to extract region proposal instead of the original Selective search method. Truly achieve end – to – end target detection framework. I didn’t really understand the importance of end-to-end until I came to understand that all models used online required end-to-end detection. I can’t say I fully understand the thinking of Faster R-CNN, and I still have a lot to learn and understand.
The real business started at the end of March of this year (prior to the demo with public data sets). At that time, due to the division of labor adjustment in the group, I began to learn from Teacher Hu Meng. From the end of March to the beginning of June, I followed up a project called prohibited Logo detection under the guidance of Teacher Hu Meng. I didn’t think so when I was running the demo, but when I got to the actual project, all kinds of problems appeared. My teacher encouraged me by saying, “It is because of these problems that we have the space to improve, and we can learn a lot in the process of solving these problems”. Indeed, from the initial model to the subsequent iterations, there is a sense of satisfaction in seeing the overall effect improve over and over again. Problems may come one after another, but the process of solving them is enjoyable and enjoyable. In this process, I found the limitations of my thinking. I needed to write the code logically and take into account all situations as much as possible. It is often the case that one or two special circumstances are missed, resulting in unsatisfactory results. I am very grateful to the teacher for timely pointing out my lack of preciseness. I feel that every time I am confused and do not know where to start, the teacher always gives me ideas to try. As the saying goes, failure is the mother of success, one more attempt, even if you fail, also one more experience of failure, slowly accumulate, thick-accumulated.
It is worth mentioning that the paper reading sharing in the group every Wednesday afternoon has always been adhered to and never stopped. Each week, a paper on computer vision is shared, usually from the latest top conference papers. Listening to the understanding of all the details in the paper, it couldn’t be better!! Many people in the group have the habit of blogging. Sometimes, I accidentally find that the blog written in detail belongs to someone in the group. I feel like the great god is around me.
The day before the end of the internship, my teacher said to me, why don’t you make a summary report of the internship work in the group before leaving tomorrow afternoon? I am very grateful for giving me this opportunity to calm down and summarize the harvest during the internship. When PREPARING the PPT, I was surprised to find that I had learned a lot and thought I had made much progress compared to when I first came.
The atmosphere in the group is really nice. The leader of the group is a little brother who graduated from Zhejiang University with a doctor’s degree. He also takes extra care of us interns. The rest of the group was also great and patiently explained when asked questions they didn’t understand. We weren’t looked down on at all because we were interns, and we were given more care. On the day before leaving the company, the team had a dinner to see me off. The first time to eat hot pot is full of meat, crayfish are dessert. At the time, I thought, what a luxury (forgive me if I haven’t seen much of the world). On the day of departure, my teacher and several colleagues in the group sent me out of the company to the bus station. Looking at the bus coming, I was suddenly very reluctant to leave. After getting along with each other for a long time, we are not just colleagues, but more like friends and relatives.
Our relationship did not end with the end of my internship, I was lucky to meet such a group of excellent young brothers. During the six months ‘internship in EDUN, I think what I gained is more than knowledge!!
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Postscript:
This intern girl is a second-year graduate student of Zhejiang University. She was originally brought by another colleague in the group. Due to business adjustment, that colleague left and transferred to my side to follow my work in March.
Strictly speaking, because she majored in control, she did not have a very deep foundation in image. At the beginning, she was really in a state of little white, but she worked very hard and seriously. She was the quiet student with excellent academic performance around us when we were students. After several months of internship, I expected her to make a summary of her work before the end of the internship, so I was surprised by the PPT at the end of the internship that she could think so much and have so many ideas (both successful experience and failed attempts are good harvest).
After the internship, she will have an internship in wechat again, and she has been thanking our team for this experience. Because of our internship experience and good training mechanism, she also easily got the internship opportunity in Tencent. As her mentor, I am very happy for her. After all, she has grown through the experience in our company and has been recognized by her peers. I think this is not only her personal recognition, but also the recognition of our team work.
Because interns positive policy this year, she after the internship, our side did not immediately give positive offer, but from her personal desire, she is very willing to join AI group continue to work in the CV related, report also said that when the netease school recruit to open in August, she will be the first time delivery on our side post, let us be there or be square ~ August
For the departing interns in the group, I have always maintained a wechat group account. With good methods and techniques, we will communicate and learn together. Again, I hope that this internship is not the end of the relationship between us, but a new starting point.
PS: The only intern girl in the group just left. We hope that more intern girls will come to our internship in the future
Finally, it is still advertising, we still recruit (social recruitment + school recruitment + internship) computer vision (face, OCR, retrieval), accelerate, optimize related to all kinds of interns/colleagues. If you are interested, please send your resume to [email protected]
Email title format: Apply for social/school recruitment/internship _(computer vision algorithm, Deep learning Optimization)(engineer, intern)_ Hangzhou Research Institute 099_ name _ highest educational background of the school.
For example: Apply for social recruitment _ computer vision algorithm engineer _ Hangzhou Research Institute 099_ Zhang SAN _ Zhejiang University master’s degree
Applied internship _ Deep learning optimization Engineer _ Hangzhou Research Institute 099_ Li Si _ Master student of Nanjing University
The name of your resume should also be consistent with the subject line of your email
This article has been published by the author Hu Meng authorized netease Cloud community, link to the original: Netease Hangyan Yi Dun internship experience (2)
Aliyun PolarDB and its Realization of Shared Storage PolarFS Technology