Python scripts for converting .json annotations from Anylabeling or LabelMe to YOLO .txt files.
-jsontoyolo.py will convert the .json annotation files to YOLO .txt files, split them into 'train' and 'validate' folders and copy over the corresponding pictures. Tracks progress using a progress bar. Requires scikit-learn and tqdm.
-jsontoyolo_simple.py will only convert the .json files to YOLO .txt files and copy them to your specified directory. Only uses built-in Python modules.
Usage:
-Define the class_labels for your dataset, example: {"car": 0, "bike": 1, "plane": 2}.
-Change input_dir and output_dir to your required directories.
-set the split_ratio, example: 0.2 # 20% of the data will go to the validation set.
-If your pictures are a different file extension then '.jpg', '.png', '.jpeg' you have to modify or add the required extension to lines: 29, 47 and 48.
-Run the script and it should generate the .txt files in the specified directory.