With Capture, four EPITA students endeavor to create smart cameras
An entrepreneurial adventure often stems from an encounter. 4th year students Augustin Abelé, Antoine Dray, Louis Marsais and Matthew Rousseau (EPITA class of 2021) met on the school’s Parisian campus in the EPITA StartUp Lab program.
In this start-up studio, the four future engineers pooled their skills, motivation, creativity and passion for artificial intelligence (AI) applied to video. What was the result of this encounter? An ambitious start-up project called Capture, whose technology was tested during the CPGE EPITA-IPSA-ESME contest, organized mainly on EPITA’s Parisian site in June, in order to ensure candidates’ social distancing during the Covid-19 crisis.
Antoine Dray, Business Developer: I was fortunate enough to enter the International Cycle when I started at EPITA, which seemed like a logical choice as I came from the Lycée International de Londres (London International High School). Before that, I had already studied a bit of pure computer science, which made me want to join the school. I’m currently majoring in Data Science and Artificial Intelligence major (SCIA). Even though we are all in the class of 2021, we have each chosen a different major!
Augustin Abelé, Software Engineer: After completing my two-year Preparatory cycle, I had the opportunity to do my semester abroad in Spain. I then decided to join the Information Systems and Software ngineering major (SIGL)… and the EPITA StartUp Lab, where I met my partners!
Louis Marsais, Systems Architect: I began my Preparatory cycle at EPITA and then did my semester abroad in Thailand. After finishing my 3rd year internship, the first in the Engineering cycle, I realized that I was quite proficient in coding, but that was not necessarily the direction I wanted to take. I therefore chose the Consulting & Innovation major (GITM) to become a consultant, a profession that holds great promise for the future! At the same time, Matthew and I decided to join the EPITA StartUpLab. Here, we joined forces with two other students, Antoine and Augustin, as all four of us wanted to work on the same subject. Besides, what interests me most about our start-up project is not the code itself, but everything else, like speaking with customers and selling a product that works.
Matthew Rousseau, Machine Learning Engineer: As for me, I joined the Preparatory cycle at EPITA as a 2nd year student, thanks to the my entry through admissions parallèles. At the time, I knew I wanted a career in IT, but wasn’t sure of the exact field. I have now joined the Image major to study AI and become specialized in the different methods of processing and analyzing images. And as I wanted to go above and beyond the classic curriculum, I also decided to apply to the EPITA StartUp Lab program. Now, the four of us have been working on a great project, in which I am more involved with the algorithm itself, which allows us to detect different events and different people!
Matthew: We learned about the lab at the beginning of our 3rd year. It was a project in addition to the classic curriculum, which could also replace the traditional end-of-studies projects in a certain way. What appealed to me was the fact that, as a student, I was able to discover how a start-up works and develop a viable entrepreneurial project.
Louis: Personally, I wanted to learn all the skills needed to develop a project and a start-up. The subject matter was irrelevant, and I was really only interested in acquiring knowledge and expertise. What appealed to me about this lab was its very concrete approach: a project is not developed simply to receive a grade, but to create a company that can remain viable after graduation. I also liked the idea of being able to do my end-of-studies internship in my own start-up. This is quite exceptional!
Antoine: Even before the creation of the EPITA StartUp Lab, I remember hearing about StartUp42, the school’s former business incubator, during my early years at EPITA. At that time, I was able to speak with alumni who had been a part of StartUp42: all of them explained that they loved the experience and spoke about it with absolute passion. But as Matthew and Louis explained, I believe that we all enrolled in EPITA because we also wanted to bring something new to society, to make a difference… And as the projects we work on for our regular curriculum are exciting, bringing a new idea and developing it is an even more appealing challenge.
Augustin: The EPITA StartUp Lab is a great place where we can put our skills into practice to develop a concrete project. It was also extremely satisfying to work on the project for the CPGE contest that was held on the school’s campus! This is exactly what I was looking for. Moreover, Antoine, Louis, Matthew and I realized that we were all very complementary, which has motivated us even more to continue our adventure.
Antoine: In the beginning, the idea was different. Initially, we were two separate groups wanting to work on AI applied to video, to process video streams. Once we got together, the first idea we started working on was to detect shoplifting in large-scale stores using AI. Finally, we realized that the market was not promising because stolen goods are not a big enough loss to use expensive smart cameras. We then changed our approach: rather than looking for a specific problem that could be resolved via surveillance cameras, we decided to provide solutions for a variety of cases that were not necessarily defined beforehand. For example, during lockdown due to Covid-19, we worked on issues regarding social distancing and wearing masks. Now, we are working on something more global that can evolve. Basically, our aim is to be able to equip surveillance cameras with artificial intelligence. In fact, needs change depending on the person and situation.
Matthew: Obviously. There are a lot of surveillance cameras used today, both in the public and private spheres. We want to provide them with intelligence. Usually, behind these cameras, there are operators simply trying to detect people or anomalies. We want to make it possible to further automate these processes.
Matthew: AI applied to video is a rapidly growing field, with the emergence of new methods for processing and analyzing video more thoroughly and much faster. This is a dynamic high-potential sector.
Antoine: It’s a field that’s not that new, however, the latest advances are huge. It would have been difficult to pinpoint a specific behavior or analyses from a set of pixels, even a few years ago. Working in this field is also about helping the human mind with projects long considered redundant. In terms of shoplifting for example, our research has shown that it is quite rare for people to continuously watch control screens, contrary to what we see in American movies. In fact, it is very complicated for someone to monitor approximately fifteen cameras at the same time. As a result, many of these cameras end up aimlessly filming, with no one watching. It is useless! That’s where Capture comes in: we can use it to provide intelligence and interesting data for each client, based on their needs.
Augustin: During lockdown, we quickly decided to focus on social distancing. Once the subject had been somewhat developed, we spoke about it with the school’s managers, including Laurent Trébulle, Director of Corporate Relations, and then everything followed fairly quickly, with a proposal from the school to conduct a test during the CPGE contest it was hosting. In addition to improving our solution, we had to obtain cameras and perform tests in order to be well-prepared when the time came. In the end, this full-scale test at the contest was clearly a great success. We were able to verify our solution, the viability of our alert system and thus obtain good statistics – which was positive for the organization of the contest as these showed that people respected social distancing rules. We also took advantage of the event to show the reporters of France 2 what we normally don’t share, namely what algorithms do on captured images. Here, individuals are framed in green or red, depending on whether or not the distance is respected.
Louis: After this test, we will work on a solution that is more in line with what we want to sell to customers in the future. For this, we plan to place the dashboard containing the collected statistics and the alert system at the heart of the solution. In addition, in terms of the software infrastructure, we are preparing the core to which we will be able to add various modules. We can add distance detection, mask detection, etc. The customer will have the choice to add any or all of these modules on his/her cameras. There will be a range of solutions to choose from.
Antoine: The idea of offering a base came from our first experience related to social distancing. In order to ensure a high performance algorithm, we had to improve individual detection, tracking – to be able to follow these people throughout the different video frames -, etc. By opting for a base, we will be able to work on a set of techniques that can then be reused depending on the different scenarios and high performance solutions that are faster to develop and deploy. The goal for us is not to start from scratch with each new customer. Moreover, we are starting to meet with potential customers.
Augustin: We have already made some contacts, thanks to the EPITA teams, as well as through LinkedIn, which drew attention to our project. Discussions are already underway to possibly use our technology in mass distribution outlets by the start of next school year.
Louis: We can also count on Daniel Jarjoura, the head of EPITA StartUp Lab, who has been coaching us and setting weekly deadlines so that our project advances. We show him what we have done, and he gives us feedback, sets goals, etc. This encourages us to work and really increase our efforts. I doubt we could have gone this far without him.
Matthew: In addition to Daniel, we also benefit from the advice given to us by those who were part of the structure in the past. This contribution is precious!
Augustin: We have the business model. It is a subscription system proportional to the number of cameras and modules selected. The subscription includes hardware installation and customer service.