Final Submission for ITU-tinyML 2023 Challenge by AI4G: Contents - Final ML model, CAD model, Final Report, Mobile Application apk(ready to install app)
Pages -
SignIn -> Home -> Map
Register -> Home -> Map
Latest Iteration
Application_demo.mp4
Preliminary Iteration - (This depects the multidevice architecture)
Wild-Life_application_demo.mp4
For the moile application I have used flutter along with google firebase - RTDB & Authentication and Bing Maps APIs to make a robust backend architecture along with a seamless frontend - user interface.
-> Firebase RTDB conducts data acquisition and analytics of the sensory and classification result values sent by the hardware-embedded application.
-> Authentication feature of Firebase allows access to registered users only and keeps a track of the activity.
-> The Bing Maps API locates the co-ordinates at which the species was spotted and marks it with a map-marker around the vicinity of the device.
For the initial casing I opted an open structure to efficiently debug the device and make possible adjustmets.
By the final iteration the exact device dimensions were determined and a protected casing has been developed.
device_demo.mp4
->The model architecture is supported by Tensorflow Lite and has proven to be highly efficient as demonstrated in the demo videos.
->It's crossplatform deployability is also evident by the seamless integration with Mobile App.
->Their is potential for a product deployment as an ebtire architecture.