Spirocount

Spirocount is a proof of concept demonstration for using deep learning to count the number of spirochetes present in an image.

This repository contains the android front-end that allows the user to load or capture images using the camera (with the help of a microscope) and retrieve a quick count of all the spirochetes appearing in the sample.

The related work on the development of the image recognition model itself is available at https://github.com/wkboyle/spirochete_vision.

Requirements

An Android device with a camera, running Android version 6.0 (Marshmallow) or higher.

Installation

1. Clone this repository and open it in android studio.
2. Build the latest model from https://github.com/wkboyle/spirochete_vision.
3. In the Spirocount project, replace app/src/main/assets/model.tflite with the model.tflite file created in step 2.
4. Use android studio to build the project and flash it onto your android device.

Usage

1. Display of the current image with each detected spirochete boxed in red.
2. Adjust the threshold, or the minimum confidence score, as rated by the image detector required to count an object as a spirochete. The number represents a percentage value between 0.0 and 1.0. The application will automatically recount the current image when this slider is adjusted.
3. Displays the number of spirochetes detected in the current image.
4. Use the device's camera to capture a new image and detect the number of spirochetes.
5. Load an image already in the device's storage and detect the number of spirochetes.

Contributing

We would love to have more images with which to improve our model. Please contact us at spirocount@gmail.com if you would like to share your images. The more images we can collect, the more we can improve the model and the quality of the image recognition.

Additionally, please submit any bug reports, feature requests, or code for the user interface by raising a github issue or pull request. Issues related to the model are also welcome, but please submit them to the appropriate repository at https://github.com/wkboyle/spirochete_vision.

Credits

This project is inspired by the Android CodeLab "Build and deploy a custom object detection model with TensorFlow Lite".

This project extends the code lab by:

• Porting the Kotlin code to Java.
• Extending image loading functionality to use device storage.
• Making image loading and scaling more robust.
• Adding user configuration of the object detection threshold.
• Displaying a count of detected objects rather than class label and confidence data.
• General refactoring and updating api usage.
• Utilizing a custom model specifically trained to detect spirochetes.