image-annotation

A python script that helps in labelling images for image classification task.

Dependencies

Make sure that you have installed:

• numpy
• opencv
• argparse

Installation

Clone and install requirements

$ git clone https://github.com/baheytharwat/image-annotation.git
$ cd image-annotation/
$ pip3 install -r requirements.txt

How it works?

$ python3 classification.py --image_folder <folder path> --num_classes  <number of classes>

For example, If the folder path you want to label its images is /home/bahey/data and the number of classes are 3. Then, The command will be:

$ python3 classification.py --image_folder /home/bahey/data --num_classes 3 

For more options :

$ python3 label.py --image_folder /home/bahey/data --num_classes 3 --output /home/bahey/data/output --start 100 --resize True 

• output: is the output path to save the labelled images. A folder called annotations is created and contains a folder to each class. default=./ (The same directory with the code)
• start: is the index of the image you want to start labelling from. default=0
• resize: If the image is not displayed well because its large size, Make it True. default=False

User manual

Accordingly to the number of classes you choose and the Images folder you specified, The images will be displayed to you and you can click:

• A number from [0 : num_classes-1] to label the image to the chosen class.
• ESC to get back to the previous image and relabel it.
• q to exit the program.
• Any other key will neglect the displayed image and not label it.