I built an app that estimates air quality using only images. The hardest part was gathering the dataset to train the CNN vision model, but I eventually got around it by using public street cameras and historic AQI data.
I also added an onboarding page to explain how this whole thing works and how it was trained. Happy with how it turns out.
It’s been a while since I last made a devlog. A lot of things have changed.
First, there’s now a onboarding section for the app. Users can learn how the model was made, how it was trained, and basically teaches you how machine learning works for vision models.
I also added a demo image page for the users to see how the model perform in harsh conditions.
Lastly, there’s also a guide page to teach you how to take the most optimal image for the model.
This will probably be my last devlog before I ship. Thank you everyone for following!
Log in to leave a comment
I decided to ditch the separate tabs and merge everything into one view. Now there’s a gallery button right on the camera screen, so you can pick photos or snap new ones without switching back and forth. It feels a lot more natural and cut out a bunch of unnecessary code.
I also added detailed guides for each air quality level, so you’ll know exactly what those numbers actually mean for your health.
Log in to leave a comment
