Joining hands to solve pressing issues businesses face nowadays, students of different universities brought their novel projects full of potential to the “Hackathon 2022 - Hack The Future” contest.
The EnerG team, composed of DTU students, won first prize at Hackathon 2022
Among them, the project “Graph neural network and computer vision app for the creation of a system to try on clothes on e-commerce platforms” by DTU youngsters won first prize in the final round.
A technology contest with real-life challenges, specifically solving difficulties or obstacles businesses meet, products with potential contesting in Hackathon rapidly receive support towards completion and real-world application.
Over 150 contestants applied, and 24 teams entered the selection round. After the contestants developed their products or projects for three months, the Hackathon 2022 final round took place on August 13 and 14, 2022, with the ten most outstanding projects with highly feasible ideas.
In the end, the organizers awarded the following prizes:
- first prize: EnerG (team of DTU students) with their “Graph neural network and computer vision app for the creation of an API for a system to try on clothes on e-commerce platforms”,
- second prize: Vn.nV (team including students from the University of Science and Technology of the University of Danang, the University of South Florida, the University of Economics Ho Chi Minh City, the University of Adelaide, the University of Economics of the University of Danang, and the National Taiwan University of Science & Technology) with their ChoIk project,
- third prize: Coachi (team of Passerelles numériques Vietnam students) with their app to “Book tickets for and manage long-distance coaches”.
As e-commerce platforms are developing, ordinary people have the possibility to buy stuff without needing to go to all the way to a shop. Going hand-in-hand with this advantage, however, is one issue frustrating businesses: return of goods when customers are not satisfied with their purchase, and particularly of clothes.
Due to this, the two DTU students of Big Data & Machine Learning Nguyen Vinh Huy and Vo Thanh Hau founded the EnerG team and came up with the idea for a “Graph neural network and computer vision app for the creation of an API for a system to try on clothes on e-commerce platforms”. The app lets customers choose the product they like, saving time and avoiding the need to return goods or try them on, which can damage them.
Using the app the DTU youngsters created is straightforward and easy. After launching the app, the user enters their height and uploads a picture of themselves. The software will draw up a first model of their bones and compute the size of their shoulders, arms, hip bones, etc.
A second model is then created to divide the picture to find the positions of different body parts, the shapes of each body part, and which pixel corresponds to which body part. A third model then sketches the clothes based on a black-and-white picture provided by the seller, after which the clothes are fitted to the body shape the customer entered.
Furthermore, this unique app allows fashion designers to create limited clothing for customers with exceptional body shapes.
EnerG team giving their presentation at the Hackathon 2022 finals
The app provides an API with a limited number of free uses and allows individual users to try on clothes on different e-commerce platforms. The app can be targeted at consumers, e-commerce platforms, shopping malls, and fashion designers.
What makes the project impressive is that it uses deep learning, computer vision, React JS, and Flask to create a system to try on clothes without touching them, fitting clothes to people connected to the GAN network, OpenPose, and LIP_JPP NET.
“At the technology side, we used React JS and Flask, a strong library and framework,” Thanh Hau explains the feasibility of their project. “Both are open source and have complete and very specific documentation and lively support communities.
“Furthermore, machine learning and deep learning are strongly developing and network architectures are constantly updated and optimized, with strong support from the ‘big guys’ in technology and with powerful frameworks such as Google’s Tensorflow and Facebook’s Pytorch.
“Based on those technologies, we completed the core of the product concerning deep learning and computer vision; the models are of good quality, and the interface has been completed to provide customers with a wonderful experience when using the product. This may be a rather novel project for Vietnam, as enterprises applying technologies like deep learning and computer vision usually only develop for themselves and do not provide their products to third parties.
“Our project is therefore meant to be shared in the community through its open API. Enterprises or individual users can use it a limited number of times for free, after which they can rent the API for a reasonable price.”
On his joy at winning the first prize of the contest, Vinh Huy says, “Through ‘Hackathon 2022 - Hack the Future’, we had an opportunity to test our skills in a large-scale competition, get experience, and engage in ‘ferocious’ competition with 23 other teams from many different universities.
“Each of the teams brought their own colors to the contest with projects full of potential, with unique ideas, combined with methodical technology elements, and very impressive presentations. We therefore feel highly honored to have won the first prize in the final round.
“We’re now planning to improve the interface and the accuracy of the model. Then we’ll integrate a chatbot and a recommendation system, to develop it into a modern and even better e-commerce page.”
(Media Center)