Abstract
The Brazilian aesthetics market is one of the most diversified and dynamic in the world, with consistent growth even in times of crisis. In 2020, even at the peak of the COVID-19 pandemic, according to a survey by the Brazilian Association of the Personal Hygiene, Perfumery and Cosmetics Industry (ABIHPEC), it had a growth of 5.8%. In addition, from 2014 to 2019, the aesthetics market grew by 567% in Brazil, occupying the fourth position in the world ranking. In these 5 years, the number of professionals in the area increased from 72,000 to more than 480,000, with emphasis on revenue in the franchise sector in the Health, Beauty and Wellness sector, which earned R$34.2 billion in 2019. In 2021, the sector regained attention when Espaçolaser, a company focused on laser hair removal treatments, launched its IPO, raising more than R$2.5 billion with its IPO. However, even in a highly competititve market as this, there is a lack of technological tools that allows follows customer journey as well as allow to precisely measure results of the procedures which the customer have been submited. Inspired by advances in Deep Learning methods, this proposal aims to study the feasibility of creating B-You, a software for mobile devices (smartphones and tablets) based on Artificial Intelligence and Computer Vision methods, which enables the collection and processing of 2D images in the wild, in order to allow different measurements of different anthropometric measurements of the body, as well as the use of such measurements to provide a better journey of monitoring results for users.To estimate the measurements, our scientific hypothesis is that the combination of the use of vision transformers, applied in the extraction of features from the images, together with the use of a multi-task learning network, where each of the output channels will be responsible for by inferring an specific anthropometric measure, it is sufficient to build an efficient and effective model for the task. In this way, we have as expected results at the end of the project, both the generation of an efficient and accurate deep learning model to perform the determination of anthropometric measurements from 2D images in the wild, as well as a software for mobile devices that allows the user to use the inference model as a cloud service, which will allow the use even in mobile devices without great computational power and low cost. From the marketing perspective, we also hope to obtain a minimum viable product at the end of the project with the potential to bring to the market a more intuitive and standardized way of monitoring the results of aesthetic procedures. It is also worth mentioning that, at the end of the project, we propose to test the product together with our partner, a network of aesthetic clinics founded in the city of Campinas-SP, which will act both in directing the construction of the UI/UX of the product, and in the collection real data to improve the models. (AU)
| Articles published in Agência FAPESP Newsletter about the research grant: |
| More itemsLess items |
| TITULO |
| Articles published in other media outlets ( ): |
| More itemsLess items |
| VEICULO: TITULO (DATA) |
| VEICULO: TITULO (DATA) |