Meet the team: interview with Computer Vision Intern Kaiyue Shen

I am a Computer Vision Intern here at Seervision, as a part of the perception team. I mainly focus on improving how we identify a subject, and how we prioritise the subjects that we track. I’m working on improving our identification processes so that we can prioritize tracking certain subjects, no matter how long they are out of frame. Other than this, I help with general development work and bug fixes where I’m needed. Like the rest of the development team, I am always working to improve and streamline the technology behind our software.

I generally like to keep up a steady rhythm of work across the day so my day isn’t very unusual. One thing that’s a bit different from the rest of the team is that I like to take a short nap as part of my lunch break. It’s pretty common in China but not so much here, so it surprised the team a bit at first. They don’t mind now, and I find it really helps to keep my energy up for the rest of the afternoon. Other than that, we have daily standups with the perception team to check in on everyone’s plan for the day and make sure we’re all on track. I wasn’t used to being in meetings so often before, but I’ve found it really interesting. We get to see what the rest of the company is working on and where the product is going.

One of our teams is currently working on a really cool project that we hope will have an impact in the sports industry. It’s in the very early stages so I can’t say much but essentially it would allow us to improve the way that we track players in team sports. Players can move very fast and maintaining tracking in these cases is an interesting field of research in camera automation and computer vision. It has a lot of potential in the industry, but as I said it’s in very early development.

Compared with doing research in a university lab, working in a company is focused on providing real world products that the customer is going to use immediately. Instead of simply running my computer vision models in a test environment, I can see the feedback from the product and interact with it directly. I can then improve the algorithm and test it directly using the live feed. Being able to see the live impact of my work is more intuitive and encouraging, especially when I see some of my algorithms really work and improve the product for the customer directly. Also, whenever I’m in the office, it’s great to meet lots of different people. Living in a foreign country during covid time makes it hard to meet people, so getting to come to the office and see a few people when I can has really helped.

I actually feel pretty lucky to have got the internship. It’s not easy as a non-European to get a work permit in Switzerland so it’s great to get to be a part of a cool company working on my own projects. I was surprised to find that there’s not too much of a gap between the academic approach and the approaches we follow here, as a lot of the necessary tasks in building computer vision models do cross over. I’m always working with different data sets and trying out different algorithms to find the solution we’re looking for. People who are much more focused around the release schedule have it a bit different to me, but it’s been great anyway to see my research being used directly by customers rather than just submitted as a test.

I’ve picked up bouldering since I’ve been in Switzerland, and go occasionally with some of the other members of the team here. Since I came to Switzerland, I’ve generally felt the fitness atmosphere, so I’ve got much more into going to the gym regularly and I take fitness and dance classes as much as I can. On weekends I like to get together with some friends to make traditional Chinese food that we miss from home, like hotpot. Sometimes we go hiking but we don’t go full Swiss (long hikes at pace), just an easy half day at a time so we’re not super tired afterwards.