Tensorflow Lite Classification App

Original repo

Implementation Details

This project is a little modified version of TFLite andorid example.

Classifier

Custom classifer can be implemented by Classier abstract class and implementing recognizeImage(Bitmap bitmap). Classifier needs to put labels file in Classifier.labels. Model (Interpretor) can be set using:

Classifier.tflite = new Interpreter(loadModelFile(assetManager, modelFilename));

Results

Classifier.Result class is used to get results from model. For showing bounding boxes, model should return bounding location of object detected.

Links

• Tensorflow Lite Github
• More TFLite examples

Requirements

Running the app

• Place model with .tflite file and labels file label.txt in 'app/src/assets' folder.
• Reference the model file in ClassifierActivity or DetectorActivity depending on type of model.
• Build app

Runtime

Android API > 21