The team from FPT Software ranked 17th on the rankings
After more than two months of competition, the FPT Augmentor team with three members who are technology engineers from FPT Software has finished at 17th place at the World Artificial Intelligence playground. In particular, two of the three members of the contest team are fresh graduates, born in 2000.
Data-Centric AI Competition, organized by DeepLearning.ai and Landing AI, is an international artificial intelligence competition initiated by Andrew Ng, who is known as the world’s AI and deep learning genius. This international competition focuses on AI training approaches based on clean data sets, optimizing artificial intelligence training time and effort, as well as bringing many advantages to businesses in many fields. join in building AI. The competition attracts nearly 500 teams from many different countries and brings together many scientific talents working in leading enterprises around the world.
Sharing about the reason for participating in the contest, Thanh Tuan, one of three members of Team FPT Augmentor, data expert of FPT Software said, “In front of an artificial intelligence project, engineers and programmers staff always have a headache to find ways to improve predictive efficiency.The question is should you try a lot of different models or focus on improving data quality, must find the optimal solution and And it’s great that one of the world’s leaders in the field of artificial intelligence – Professor Andrew Ng creates a large and extremely professional playing field.”
With the Data-Centric AI Competition, the teams will receive questions related to the data set of letters in roman notation. Players need to use data processing techniques, submit photo folders of no more than 10,000 photos and calculate the results based on accuracy. During the competition, data engineers are required to make decisions and analyze technical specifications, thereby coming up with changes and solutions to improve the quality of the original dataset. head.
Three talented members of the competition team from FPT Software
“When we ran the test, we realized that the model hadn’t learned the difficult images, so we picked out each layer the difficult and strange images to focus on enhancing these images and got an 84.256% result.” – Tuan shared more about the process of enrolling in the top group on the “yellow board”.
Inspired by the above international competition, in Vietnam, FPT Software also hosts for the first time the Data-Centric AI Competition in 2021. The topic of the contest in Vietnam has the theme “Wearing a mask” page” with a dataset of 976 images labeled in 3 classes: wearing a mask, not wearing a mask, and wearing the wrong mask.
The organizers provide a set of trained models and a public test set of 89 images so that candidates can make a preliminary assessment of the model, in order to find smart solutions in the prevention of the Covid-19 pandemic. The contestants participating in the contest were guided by the world’s leading AI experts: Prof Andrew Ng, Stanford University Lecturer – Huyen Chip, Google – Huu Tiep,… Readers can see more information at Link
At FPT Software, the data-centric approach is also widely applied in Artificial Intelligence projects, helping experts easily handle problems with small to large data sets. in many different industries such as aviation, medical, finance, insurance without changing too much in the model. This approach saves engineers time and increases project efficiency and accuracy.