When we first heard about the theme of mental health, there were a lot of different subtopics and possible solutions that came to our mind, but a lot of them already existed and we didn't see anything exciting and special about them. However, we had some experience with machine learning and were really enthusiastic about it, so we came up with the idea of PillChill. Many people have a lot of random pill bottles in the medicine cabinet and they have no idea how long they have been there, what they actually do, etc. Not to mention the fact that sometimes people will reuse old pill bottles, making things even more confusing. We realized that people with mental health issues as well as other health issues might have a hard time remembering what meds to take each day and when to take them. In addition, people who take anti-depressants and anxiety meds would also have challenges while trying to wean off of them safely and effectively.
What it does
PillChill is an app that uses machine learning software to identify pills from an image that the user takes and directs them to a website that gives the user more information about the pill, so that they can make sure that they know what they are taking. Another feature of PillChill is the schedule page, where the user can input their schedule for their medication. It also has a page for alternative remedies, so if the user is struggling with their mental health and needs some additional support along with their prescribed medication, they have access to those resources. The alternative remedies page has a section for ranting, which directs the user to a note page where they can let all their emotions out, which can be very therapeutic.
How we built it
We used Tensor Flow, which uses Python, for the machine learning aspect of the app for image recognition. We built a prototype of the app using BuildFire, but have started development of an actual Android app in Android Studio.
Challenges we ran into
We are not experts at machine learning, or app development, so as we were coding, we ran into problems where certain things weren't working like we thought they should. It took some time to figure out how to tune the machine learning to recognize pills and write code that was better suited for identifying pills. We also had trouble trying to get the buttons to align properly on the app and we had to do some research and reading to learn how to connect the phone's camera to the app. However, we worked through these challenges, tried different things, and eventually learned more and became a better programmer through these struggles.
Accomplishments that we're proud of
We are especially proud of the Tensor Flow because machine learning is something that we are very interested in so getting it to work was very exciting. We are also proud of the nice design for the prototype of our app and are working on implementing that in our Android app.
What we learned
We gained a stronger understanding of machine learning and how to use Tensor Flow. We also became more comfortable with using Android Studio and learned more about how to add functions to buttons and link all the pages together.
What's next for PillChill
We want to code the app so that the user will get notification reminders of every time they need to take their meds. The notifications would sync up with the schedule that the user inputs. We also want to add a feature that allows users to plan out how they are going to try and wean off of medication, with input from their clinician. This would be integrated with the schedule page of the app.