Handwritten-Digit-Recognition-on-Android-Mobile

This project is based on recognition of single hamdwritten digit 0 to 9 using android app created in Android-Studio.

Here is several steps to run this whole project.

1. Install dependancies.

Install tensorflow 1.2.1 (also need to install anaconda for python 2.7)

Install other dependancies:

• install csv (pip install csv)
• install opencv (pip install opencv-python)
• install numpy (pip install numpy)

2. Download dataset from Site.

Donwload dataset for Training images

Download dataset for Training Labels

Download dataset for Test Images

Download dataset for Test Labels

3. Run python files for training.

First, Run python file.

python mnist2jpg.py

Then run another python file.

python train_mnist.py

4. Build android project and run app on any android device(>lollipop)

First install android studio and config sdk with software as per software suggest. After compliting all configuration,import project from tfmobile. Import output_graph.pb and output_labels.txt which have created in out/ when one run python train_mnist.py in tfmobile/asserts. Change in ClassificationActivity.java which is in tfmobile/src/org/tensorflow/demo. Changes are given below

private static final int INPUT_SIZE = 28; private static final int IMAGE_MEAN = 128; private static final float IMAGE_STD = 128.0f; private static final String INPUT_NAME = "input"; private static final String OUTPUT_NAME = "output";

Then Build and Run OR check tfmobile/gradleBuild/outputs/apk/debug and get tfmobile-debug.apk to run on any adroid devices(>lollipop).

If is there any issue in making app one can take reference from application reference.