Android Chicken Prediction Bot

I want to turn the quirky idea of a “Chicken Road Prediction Bot” into a fully working Android app. The concept is simple: feed the app either live video or pre-recorded movement data of chickens near a roadway and have it predict if, when, and where a bird will step onto the road. The platform is strictly mobile and Android is the only operating system I need supported.

Here’s what I expect:
• A lightweight Android application (Kotlin or Java) with a clean UI that shows a camera preview or data feed and overlays the real-time prediction.
• An on-device model—TensorFlow Lite, PyTorch Mobile, or a comparable framework—that can run offline and outputs a confidence score plus an ETA for the crossing.
• A simple stats panel logging every prediction with time-stamp so I can review accuracy.
• Full source code, model files, and build instructions so I can re-train or tweak the logic later.

I’ll provide any chicken-movement footage I already have, but please plan on suggesting additional sample-collection or data-augmentation strategies if the model needs more variety. The project is done when the APK installs on a modern Android phone, processes a short test clip in real time, and consistently flags crossings at a usable accuracy level.