Skip to content

Latest commit

 

History

History
52 lines (36 loc) · 3.93 KB

README.md

File metadata and controls

52 lines (36 loc) · 3.93 KB

alt JibJib

Repo layout

This is the repo for our Android app.

The complete list of JibJib repos is:

  • jibjib: Our Android app. Records sounds and looks fantastic.
  • deploy: Instructions to deploy the JibJib stack.
  • jibjib-model: Code for training the machine learning model for bird classification
  • jibjib-api: Main API to receive database requests & audio files.
  • jibjib-data: A MongoDB instance holding information about detectable birds.
  • jibjib-query: A thin Python Flask API that handles communication with the TensorFlow Serving instance.
  • gopeana: A API client for Europeana, written in Go.
  • voice-grabber: A collection of scripts to construct the dataset required for model training

An Android app for matching bird sounds

JibJib is a mobile App for Android, which provides access to ornithology in a clear, minimalistic and playful manner. The app focuses on an intuitive and clear design and workflow to guide the user through the different use cases.

Getting started

  1. Clone the repository
  2. Create a new file ''gradle.properties'' in the root directory of the project.
  3. Add the property ''WEBServiceBaseURL = <url_of_the_backend>'' with the URL of the backend service.

For providing your own backend service, take a look at the following repositories:

Note: You can enter some dummy URL first to be able to compile and use the app. But of course, trying to match your recorded sound will result in an error message. "Oops, something went wrong"

Application workflow

alt Application workflow

When hearing a bird call the user can use his Android device to record the sound of the call. The recorded audio data is sent to the REST API where it is processed and match result with the corresponding matching accuracies are returned. On the following screen the match results are displayed comprising the scientific and common name of the bird and the matching accuracy. Each of the results can be clicked on to show more detailed information about the bird. Store bird to local data base On the detail view it is also possible to add the bird to the collection i.e. storing the bird into the local app data base. The local bird collection can also be viewed separately.

Clean Architecture

alt Clean architecture

The app is built using the ''clean'' architecture. This architecture divides the software into different modules by means of separation of concerns. This provides a maximum of loose coupling between the different modules. The diagram visualizes how the architecture works. From the outer to the inner layer the implementation gets more abstract. The inwards dependency rule states that no inner ring has any dependencies of any outer ring. The core modules contains the business entities i.e. the objects the app works with: audio files, birds, match results etc. In the layer ''above'' the use cases or functionalities of the app are defined. It is important to note that these two layers are completely independent of any framework meaning they don't even "know" anything Android related. The outer layers contain concrete implementations of frameworks, UI, data base and web access, etc. This modularization defines a sustainable architecture which makes it possible to test each layer independently from the other layers and also replace any layer without a huge impact on the rest of the code base.