This project was developed as the final project for the subject of mobile device programming (CC3035). The objective of this project (project 3) is to implement an application to identity different foods in a place/plate. The application is be able to perform real-time analysis of a video stream and track different foods on a plate as well as calculating its calories, in this application you can also save and visualize all your detected foods if they are saved in a post.
This application is supported by a Django backend and frontend server/site where you can also download the app https://178.79.132.23/ .
To recognize the foods, it uses our custom built dataset with 90 classes of foods and uses a pretrained SSD model that trains on our data using the TensorFlow library (the notebook used to train the model can be seen in the Model folder). Since we want to integrate our model with the app, Tensorflow 2.0 API now supports a new model format (tflite) that can reduce the consuption of resources during inference, as such, after training the model we convert it to a tflite format.
In case you want to download and test the app you can download it in our website or in this repositorty or even in our Aptoide store. Since the app is not deployed on PlayStore you will need to enable unknown sources and allow the installation on your device, you can find out how to use the application here, to use the full features of the app you will need to authenticate, consenting to our terms and conditions. If you want a better look into the development process you can check out our project here.
If you are using this app as reference, you can check our dependency gradle file here and the app graddle here.
Note: This project's server is no longer running, as such, all website links won't work. The REST interaction between the app and the server will also not work, meaning that, all actions that require server communication in the app (login, sign up, all operations insider logged session) will no longer be supported in the app, you will now only have access to the no authentication functionality.
If you want to see the whole demo go here.
