Update README.md

README.md

+Name : Gunasinghe M.D.
+
+IT Number : IT17043342
+
+Group ID : 2020-046
+
+Research : CIAT - Child Intelligent Assessment Tool
+
+Individual Component: SCIT - Shapes and Color Identification Test
+
+
+
+**Please read the instructions section to train the cnn model and run the application** 
+
+
 # Shapes and Color Identification Test
-**Abstract**
+**Abstract** / **Research Question**
 
 The shapes and colour identification test conducted manually in Lady Ridgeway Hospital in Sri Lanka . Since there are issues with manual data storing and the accuracy of the test results , as a solution we have proposed an automated application. The SCIT will help doctors and parents to identify the learning disabilities of preschool children.
 
@@ -34,3 +49,67 @@ In this test user will be ordered to copy the shapes on canvas which is provided
 Generated accuracy as a percentage of actual drawn shape on the canvas. Analysing the results according to factors such as gender , age difference , IQ level and the way drawing the shapes.
 
 
+
+# project structure
+
+
+
+*  "CNNmodel" : folder includes python scripts for generate the cnn model
+
+*  "ShapesClassification" : includes android studio project for shapes identification app
+
+
+
+# Instructions - Model Training /CNNmodel
+
+
+As a first step i trained the model using training dataset within 3 main classes.
+
+After that save my trained model into a h5 file.
+
+i've used this pretrained model to predict our custom images(single) from test dataset.
+
+at the end of the training process visulizing accuracy and loss of validation after each epoch using matplotlib library.
+
+
+
+**please install requirements - python libraries**
+
+    $ pip install -r requirements.txt
+    
+    
+to train the model using training dataset    
+
+    $ python cnn.py
+
+
+To give single image of shape stored on your computer and predict use script as below
+
+    $ python predict.py --image <image-path>
+    
+You can find the confusion matrix and accuracy of model after prediction.
+
+
+**The Model Conversion and Integration Task**
+
+
+after the training process you can simply run "convert.py" to convert keras model(.h5) into (.tflite)model.
+
+
+# Instructions - Android Application 
+
+    Clone the this repository -> open in android studio -> run the application 
+
+Android studio will download necessary plugins to the environment automatically. after that you can run this application.
+
+also **i've provided apk file** so you can simply copy that into your mobile phone and install and run it.
+
+
+
+# System Environment
+
+    windows 10
+    Android Studio 4.0.0
+    Python 3.7
+    Keras 2.2.0
+    Tensorflow 2.2.0rc4 
\ No newline at end of file