Skip to content

Pytorch based Android app, which classifies images using MobileNet-V2 model, takes image using CameraX API

License

Notifications You must be signed in to change notification settings

darkmatter18/WhatsThere

Repository files navigation

WhatsThere

Installation Guideline

GitHub Forks GitHub issues GitHub stars GitHub license

App Splash Screen

An Android Application, which can sense and classify objects in Real time.

The App is created using the concepts of Deep Learning and Artificial Neural Networks. WhatsThere is a great example, that how Artificial Intelligence can do a lot without human interference.

Features

Features

  1. The App can sense 1000 different kinds of objects, from Keyborads to Dog Bread
  2. The App has a Flash On/Off button, so the app can see clearly at night.
  3. The Camera Focus is automatically adjusted, so you don't have to bother to focus any object.

Last but not least

  1. The App is so light weight and can do it's job very precisely even in VERY LOW SPECS Device.

Technology

Android PyTorch JAVA

  • The App is created for and only for Android, 2 minutes of silence for other platforms
  • To power-up the Deep Neural Network, the App uses PyTorch
  • Written in Shuddh JAVA
  • Uses CameraX to Capture frames

Neural Network

  • The App uses the pretrained MobileNet-v2 model as it's classifier backend

Model Artitecture

Example Images

App Example 1

Maintainer

Relative date Maintenance

Arkadip Bhattacharya

Need help? Feel free to contact me @ in2arkadipb13@gmail.com

GitHub followers Twitter Follow

LOVE FOR YOU BY DEVELOPERS