imageAI

This repository contains basic and understandable level of a neural network example. You can understand the fundamental working mechanism of machine learning concept. Packages used in this project:

• tensorflow
• keras
• matplotlib
• numpy
• cv2
• pickle
• random
• os

You can install packages by using the requirements text file:

pip install -r requirements.txt

Process Data

dataOrganize.py will check your data and extract the essential data from them.

You need to create and classify your folders by considering your class names.

Example :

• DataFolder
  • classOne
  • classTwo
  • classThree
  • empty

empty folder will contain unclassified elements.

Insert your data folder. Example:

DATADIR = ".\DataFolder"

Create Neural Network

This chunk of code optimizes your model. Different kind of settings can be performed by changing arguments.

history = model.fit(X, y, batch_size=64, epochs=40, validation_split=0.1)

In this line you can change the configs according your usage.

Testing The Model With Images

testWithImg.py used to test images.

CATEGORIES = ["classOne", "classTwo", "classThree", "empty"] # categories you have. Should be in same order...

The CATEGORIES list item names should be in same order otherwise you can't see the true class name.

image = "path-to-sample-image" # image of sample

In this line you should define path of your test image.

TIPS

Windows uses the backslash "" for seperating path elements in Python it may escape the next character so you need to use double backslash "\"

C:\\Users\\User\\Folder\\File

or

/home/user/folder/file

Note: Different kind of solutions for path problem can be used.

Don't forget the use same model number on your operations

the name that defined in this line

model = tf.keras.models.load_model("MODEL_NAME.model")

should be same with :

model.save('MODEL_NAME.model') # this line should be same with testWithImg.py