-
-
Home Page
-
Detecting unsafe posture
-
Detecting safe posture
Inspiration
While Computer Vision based Exercise Assistants are not uncommon, I realized I had never seen such a platform specifically designed for pregnant women. After a little bit of research I found out that during pregnancy, staying active with the right exercises can be extremely beneficial, but only when done safely. Many pregnant women are unsure whether their posture during common workouts like squats, lunges, or yoga-based movements is safe for their stage of pregnancy. So I decided to take matters into my own hands and build a small prototype of a platform that can grow much bigger with the right guidance and mentorship.
What it does
BumpBot is an AI-powered web app that helps pregnant women exercise safely by: 1) Using a webcam to detect real-time posture. 2) Analyzing common exercises like Squats and Bird Dog. 3) Users can select their trimester and the body angle threshold changes accordingly (catering to the bump size) 4) Giving instant visual feedback on whether the pose is safe or unsafe.
How I built it
I built BumpBot using the following tools and technologies: 1) Streamlit for creating the interactive web app interface. 2) OpenCV to access and process webcam video in real time. 3) MediaPipe for detecting body landmarks and analyzing posture. 4) NumPy for angle calculations between key joints. 5) Python as the core language tying everything together.
Challenges I ran into
1) Since this app is tailored for pregnant women, I couldn’t easily test it with the exact target audience, so I had to use some images for reference as you can see in the video. 2) Figuring out the safe angle thresholds for each trimester required a lot of research and it was tricky to do within a limited time
Accomplishments that I am proud of
I have had this idea in mind since a year now, and I am so grateful and proud that I was finally able to implement it. Although, it does not contain everything that I had in mind, I am proud and impressed by how I could come up with the logic ( with some help from online resources, of course) and design it within such a short time.
What I learned
1) I gained hands-on experience with MediaPipe and learned how to interpret landmark data for exercise form analysis. 2) I also learned how Streamlit can be used for handling sessions, layouts and moving smoothly from one page to another. 3) Mainly, I learned how tech can be used to improve the lives of so many people, especially in the field of healthcare. Tech must adapt to the user, not the other way around.
What's next for BumpBot
Here's how I see BumpBot in the future: 1) A mobile-friendly platform with voice instructions and safety alerts for hands-free workouts. 2) An option for healthcare professionals to validate users' movements and have one-on-one sessions with them. 3) Multi-Language Options. 4) Track reps and store exercise data which can be displayed on an interactive dashboard.
Log in or sign up for Devpost to join the conversation.