Preprocessing images - Annotation

This ReadMe explains how to use this tool to do annotation

Features include:

Do annotation and export the file under Darknet and Pytorch format
Resize the images

Getting Started

In order to setup the project, please follow the next steps.

Copy the tool

Copy the tool from:

EFR-DL0063-IDEE Data Services - General\40 - Microservices\417 - MS28 Automatic element detection in pictures\4173 - Investigation\R&D_MS28_V1\labelImg

Prerequisites

python 3.7

Requirements

pip install -r requirements.txt

You can define your label list at:

.\data\predefined_classes.txt

Run the tool to do annotation

# At terminal:
python labelImg.py

Note: Pay attention at the directory where you save annotated images, make sur that there are at least two exported files .txt for each image, one is _pytorch.txt If not, please save at least two times while doing annotation for one image

Convert the dataset from Darknet format to Pytorch format

Open dir.txt and copy this link to full_path_to_images in creating_train_and_test_txt_files.py.
Define the ratio of train/test (default is 15% of test, 75% of training)
Modify the appropriate size of images to do resizing (default is 416x416)

# At terminal:
python creating_train_and_test_txt_files.py

TrungKien1230/labelImg_pytorch