On October 20, 2021, in ACM Multimedia 2021, the multi-mode commodity recognition Workshop and the second Taobao Live Commodity Recognition Contest jointly held by Alibaba Tao Department technology and Zhejiang University was successfully concluded and the prize was awarded onsite.
We talked to Gong Litong, the first winner from the Institute of Computing Science of the Chinese Academy of Sciences. The competition gave him a perspective to understand the business of the industry, and we were able to better understand the technological dreams of the young people who are coming up. ‘.
As a competitive teenager who plays at least three algorithm competitions each semester, Gong’s goal is simple: he forces himself to learn new things.
At about the age of three, when he secretly took apart his remote control car and put the circuit board back on, he discovered his first truism — practice makes sense.
Not long ago, he is still in the Computing institute of the Chinese Academy of Sciences, he won the tao department of technology and Zhejiang University jointly held the second session of Taobao live commodity identification contest first place, and he will feel after talking, the first place is really not unexpected. 🙂
“When I was playing, I accidentally bought a dress on Taobao live broadcast.”
The topic of this competition is multi-modal commodity retrieval and recognition based on the understanding of taobao live data content. Compared with other contestants who started to clean up data, make models, adjust parameters and train machines after they got the questions, an engineering teenager who only spent his spare time in the lab checking digital bloggers, actually started to check live on Taobao.
Gong Li copper thought, since this is a business type of data set, always want to see what business is doing.
He switched to the channel of clothing live broadcast and carefully felt the clothing content that the owner wanted to show and the corresponding language, especially the correlation between intention and visual information. After watching several shows, he finally succeeded in being planted and bought a piece of clothing…
Combined with the perception and understanding of various specific types of lighting in the broadcast room, clothing display forms, even in the face of word segmentation text marked by numbers, Gong Ligong obviously has a clearer business logic and algorithm thinking, and directly bypats some detours in model building and detail processing.
On July 20, when the results were set, The F1 score of Gong Li Tong’s single team was as high as 0.69, exceeding the baseline of 0.22 and winning the safe champion.
“I have been quiet and thoughtful since I was a child.”
Gong Litong was born in Zibo, Shandong province, the former capital of qi. Under the influence of the culture at the origin of Cuju, he liked Messi and football, but he seemed to prefer playing with his toys and games quietly and skillfully than playing with his friends.
The father that does engineering technician in building company, the toy that buys to him also has very “professional style”, kongming lock, rubik’s cube, remote control car, jigsaw puzzle goes all out to assemble a kind to wait a moment, tore open recombination in infinite round, group tore open again in polish, gong li copper feels also “more and more top”.
When I graduated from primary school, my family got their first computer. Different from the parents of the game “beast of the flood” general attitude, dad will invite Gong Li Copper to play the game, red alarm, legend, adventure island… Every children in that era to secretly run to Internet cafes to get happy, Gong Li copper and dad together to complete the practice.
“I felt like it was his’ scheme ‘because I never got hooked.”
Without this “forbidden zone temptation”, Gong li Tong’s simple joy has always been on how to get more ingenious results through thinking.
Math had always been a high score, and he wasn’t satisfied. Disdaying the methods that can be thought of at a glance, Gong always tries to find a more sophisticated Angle. Math and physics teacher often speechless, the child how always contrary to the original intention of the question, but the answer is very correct……
A monthly exam known as “the most difficult in history”, the first place gong Li copper 117 points (full score of 120), the second student is less than 100. The teacher decided that this kind of thinking is not very conventional examination, later not whole.
Like all teenagers, he adored Einstein and hoped that one day he, too, would break through “common sense” and “rules” to challenge the more fundamental problems, the bigger problems, the ones that have not been solved, the ones that don’t know how to solve.
“I probably want to take apart some of the smart devices around me. I’m pretty sure about my professional direction.”
Drones, smart cars, electronic stopwatches, bus voice-reading systems, automation equipment on factory lines… Everything in this world according to some rules of writing in the automatic operation of things, Gong Li tong has a curious impulse to explore.
It seemed that he had found the entrance to a certain life. Self-discovery and confirmation in his youth made him never hesitate and confused about his major choice.
When he graduated from senior three, he applied for the electronic information major of Shandong University. In the four years since then, he has been engaged in the innovation and entrepreneurship competition, making small and large pieces such as electronic blood pressure monitor, “one-hop” physical plug, and intelligent car that can automatically go through the maze.
In the second semester of his junior year, he was enrolled in chongxin Academy, an experimental engineering class that advocates innovation and hands-on practice. At that time, there was a topic [intelligent car to automatically run the maze]. The car used for the experiment had been discontinued abroad, and there were nearly 30 students who needed to use the only car.
So what to do? Gong Li copper and roommate spent half a year, actually made a car performance completely close to the same.
Memories of taking apart remote control cars as a child came flooding back, and he could feel his every nerve firing. The vehicle, which can avoid obstacles automatically, has extremely high awareness systems and sophisticated controls, making it almost impossible to replicate the same performance without factory support. Gong bought gyroscopes, sonar, Bluetooth connectors and other parts from Taobao and tried to write drivers and put them together. And then… He blew up, the ports burned, parts failed, and a pile of iron sheets stood there, as if to announce a cool end.
Volume soil again, Gong Li copper carefully review the entire operating system, determine sonar is the most critical and complex place, the need for real-time calculation of all perceived distance and obstacles to internal computing systems and communication protocols connected to build intelligent car driving map. Under the support of the high-precision sonar measurement equipment provided by the teacher, they step by step, slowly and steadily, small debugging, and finally really imitation of a similar intelligent car.
Learning by doing is the best way he has always pursued. Instead of getting a car made by himself, his efforts to consult, debug, write and so on to solve each specific problem in this process can help him to shape his knowledge system.
“You can’t know exactly what the problem is until you’ve experienced it.”
Gong Litong chose CV direction when he went to institute of Computing Science of Chinese Academy of Sciences. For the gradual development of experience for him, graduate algorithm is a new challenge. Of course, the more realistic reason is that in 2017, he made a recruitment information database for a big assignment of a course. After he finished crawling all kinds of job information such as Java, NLP and CV, he smelled the current situation of artificial intelligence blowout more keenly than his fellow developers.
For the development of just turn algorithm friends, the common problem is how to learn and get started. For Gong Litong, a graduate student, it takes at least 1 week to understand a paper, and another week to help the project in the group match the environment. On one of his most anxious days, he rushed to the lab at 3am and couldn’t sleep until 11am the next morning to get the environment configured. “I can’t help it. I’m not familiar with each module. I can only feel it after practice.”
From there, he fell in love with the game. Each question usually represents a specific knowledge practice in a certain field. Gong will throw himself into the competition to study the context and the latest progress of this direction, and constantly test the model and practice after completing the previous knowledge learning in books and videos. Throughout graduate school, he played in more than 10 algorithm contests. The result of the competition is not important, he learned the knowledge and practice of image retrieval and classification, object detection, OCR and other aspects in this way.
He has been trying to summarize and review his experience and winning points, “such as forming a description or an algorithm, focusing on the essential logic and refining a set of experience in analyzing and solving problems by himself; I also try to make myself understood.
Gong Ligong does not have any internship or work experience, as a student, he has been climbing the peak of knowledge, carefully polishing the basic practical skills. Most of his competitions are individual, and teams are rarely formed, because individual races offer greater freedom and space for learning and exploration.
It is very commendable that he still has a clear understanding of the challenges he will face in the future work. He is not an excellent student blindly immersed in his achievements and learning.
Without actual work experience, he will have a time to hone his ux and business-oriented perspective.
Lack of team work experience, he will need to try more communication and integration to achieve common goals, rather than just complete a set of sub-tasks.
This is a teenager who “does what he does at what stage” and “can see the next stage”, so he does not have much of the confused anxiety of his peers. In a pile of autumn offers, there are clear selection criteria and methods.
“I haven’t decided where yet. However, I pay more attention to the growth of the technology and the core of the business, so that I can use my abilities and strengths.
Of course, because his naive girlfriend is also doing technology development in Beijing, Gong Litong will also stay in the city more than 2 hours from his hometown. Beijing is so bright in autumn, and the maple leaves in Fragrant Hills are their favorite colors.
“In the future, I hope to be able to do technology that has real business value and can serve society.”
Working on projects in the lab, mostly purely academic and clean data sets, comes and goes thinking about how to find better models and improve points. For students in school, Gong Litong felt that the taobao live commodity identification competition can bring his business thinking, is very meaningful.
“Industry data sets were linked to real-world scenarios, such as the identification of anchor intentions in this live stream, which increased the excitement of the task. (^ del ^)”
In the mind of this young man with a grand technological dream, the value of technology should finally be reflected through business, and he can really use innovation and progress to promote the progress of society. There are always moments when we wonder if there is a more perfect answer to our present life. It may be a technical problem, it may be a mathematical problem, it may be the basic science that underlies the development of our social life.
Values can be very big. Happiness can be very small.
At 12 o ‘clock at night, and students after watching the movie “I and my hometown” on the way back to the dormitory, Gong Li-tong’s mind emerged today to see a mobile phone experience store, suddenly to get a better model to the heart. He rushed back to the lab, clanged until 3am, and found that “it worked”, “it worked”, “the data was better”!
For technical people, this kind of cool feeling, more than all the happiness.