Inspiration
We chose to focus on this topic because most fitness apps, although advertised to all genders, use ML algorithms that are trained with primarily male data. For that reason, there is a prevailing gender bias in AI algorithms. This is related to the fact that medical data has been male dominated for years. An article by the Stanford Social Innovation Review, similarly claimed that women are not included in medical trials since its claimed that their bodies are too “variable and complex” (Smith & Rustagi, “When Good Algorithms Go Sexist: Why and How to Advance AI Gender Equity”).
What it does
The idea for our project is a fitness app that is catered towards women and gender minorities. It focuses on using factors like height, weight, menstrual cycle, age, and level of physical activity in one’s lifestyle. Using this information, the system provides the user with exercise tips and fitness goals for different motivations. For instance, if a user wants to lose weight, gain muscle content, or just increase daily activity, the app will provide a plan that the user can follow.
How we built it
We found datasets on Kaggle related to exercise styles and calorie loss with only female samples, data on menstrual cycles, average female height and weight and other medical datasets with only female samples. We then started to build a machine learning algorithm that trained on the following datasets using Pytorch. We also drew up the design for our app.
Challenges we ran into
Initially, we wanted to build a prototype for our envisioned project, however, there were several time and resource constraints. For instance, we were not able to find enough datasets that pertained to our topic as well as the issue of installing several applications to create an AI model.
Accomplishments that we're proud of
Despite the setbacks we faced during the hackathon we managed to come up with a good proof-of-concept.
What we learned
We had not used Pytorch yet as a machine learning software so it was interesting to get comfortable with it, rather than using TensorFlow. Additionally, we learned more about how to create machine-learning algorithms and recommendation models.
What's next for Cycle
Our envision of this project for the future would be to incorporate tools that would allow the user to provide feedback about whether the AI's suggestions are effective. This would help improve the app as well as be useful if incorporated into the healthcare system to provide helpful solutions to patients.
Log in or sign up for Devpost to join the conversation.