What I've been through in a year / 私が1年間で経験してきたこと

My name is Takumi Owada and I have been working as an intern at Orange TechLab.While I was doing my research as a graduate student in image processing, I had a chance to work as an intern with Miyazaki, the president of the company. In the past year, I have learned not only about Deep Learning, but also about quantum computers and many other things, and it has given me a lot of strength for the future of my working life. I would like to talk about the experience I've had. At the beginning of my internship, I only had a basic understanding of the theory of Deep Learning, and I had never actually put it into operation, let alone understood what kind of networks were hot at the moment. As a beginner, I was introduced to the various object detectors and was carefully instructed on how to operate them, and I was provided with an environment in which I could move them. My main activity was to improve the accuracy of re-ID and tracking using DeepCC, a method of MTMCT. I did some thesis research and open source research to achieve this goal, but I had the hardest time doing follow-up experiments. It was a storm of errors to install MATLAB, which I am not familiar with, and gcc, which is specified as a version, but it finally worked with vanilla at the end of the year. Under normal circumstances, we should have done this until we applied the experimental data held in the orange tech lab to DeepCC, but we had to give it up halfway through the process due to a series of errors. It was very frustrating to take the time to build the environment, but it was good to gain a better understanding of Deep Learning and how to deal with errors.I will also be accompanying the CVPR, one of the largest conferences in the world, as part of the MMTMCT's research. It was the first time for me to attend an international conference and the first time abroad, and I gained a lot of experience not only in the latest research of MTMCT, but also in the knowledge of the research I am doing as a graduate student. I would like to thank Ms. Tashiro and Dr. Nishida for taking care of me during this trip. During this internship, I had the opportunity to learn about quantum computers apart from Deep Learning. I was interested in what was different from the computers we are dealing with now, and how they were implemented, so I attended more than a dozen independent seminars that lasted about two hours. My impressions of the seminar were that it was completely different from what I had imagined from what I had seen and heard on the news, and that it was difficult for me to understand enough to explain it to others, even though I had participated in the seminar many times. This experience has given me a certain amount of dispassionate view of quantum computer-related news. I would like to thank Mr. Asakura and Mr. Akira Miyazaki for the opportunity to learn about quantum computers, which will become a hot word in the future.

Translated with www.DeepL.com/Translator (free version)


私は画像処理系の大学院生として研究を行っていた傍ら、社長である宮崎社長との間にご縁が生まれ、インターンをさせていただく運びになりました。この1年でDeep Learningの知識だけでなく、量子コンピュータやその他多くのことについて学ばせていただき、これからの私の社会人人生に多くの力を与えてくれました。これから私が得られた経験についてお話ししたいと思います。

インターン開始当初はDeep Learningに関して理論のさわりしか把握しておらず、どのようなネットワークが今アツいのか把握していることはおろか、実際に動作させたこともありませんでした。そのような初心者の私に、様々な物体検出器の紹介とその動作方法について丁寧に指導いただき、また動かせる環境を提供していただきました。私の主な活動はMTMCTの一手法であるDeepCCを用いてRe-IDとtrackingの精度向上を目指すことでした。この目的を達成するために論文調査やオープンソースの調査を行ったのですが、追実験を行うことに一番苦労しました。使い慣れないMATLAB、version指定されているgccをインストールするのにもエラーの嵐で、ようやくバニラで動いたのは年末でした。本来であれば、オレンジテクラボで保有している実験データをDeepCCに適用するまで行うはずでしたが、こちらもエラーの連続で道半ばにして断念することになりました。環境構築で手間取ったことは大変悔しいですが、Deep Learningに関する知識、エラー対処について理解を深めることが出来たのは良かったです。


このインターン期間中、Deep Learningとは別に量子コンピュータについて勉強する機会がありました。我々が今取り扱っているコンピュータと何が異なるのか、どのような実装をしているのかに興味があり、2時間ほどの自主ゼミに十数回参加しました。参加した感想としては、自分が今までニュースで見聞きして想像してたものとは全然異なるということと、何回もゼミに参加してる割には、人に説明できるほど理解できず難しかったということでした。この経験で、ある程度量子コンピュータ関連のニュースを冷静に見ることが出来ました。今後、ホットワードになっていく量子コンピュータについて知ることが出来た機会を下さった朝倉様、宮崎彬様に感謝申し上げます。