labelme_yolo_utils

Utils for labelme and YOLOv5 detector

convert_labelme2yolo

Convert LabelMe JSON to YOLO txt.

example:

./convert_labelme2yolo.py --input=./photos --output=./res --classes=./class_names.txt

convert_yolo2labelme

Convert YOLO txt labels to LabelMe JSON.

example:

./convert_yolo2labelme.py --input=./res --output=./res_json --classes=./res/class_names.txt

Takes data from input path: read images from "images" and labels from "labels" directories.

Example:

    ./input_dir
    |-> /images/*.jpg
    `-> /labels/*.txt

predetect_yolo2labelme

Make detection on image and store results into LabelMe JSON format for next manual labeling. Could be useful for processes of Semi-supervised learning or Active Learning.

example:

./predetect_yolo2labelme.py --input=./photos/ --model=./yolov5s.pt --classes=./coco_class_names.txt --threshold=0.3

copy_unlabeled_images

Copy (or move) images without LabelMe JSON files to output directory.

example:

# copy
copy_unlabeled_images.py --input=./images --output=./unlabeled --extention="jpg"
# move 
copy_unlabeled_images.py --input=./images --output=./unlabeled --extention="jpg" --move

cut_image_from_labelme

Cut image from LabelMe JSON bounding box to image file.

example:

./cut_image_from_labelme.py --input=./images --output=./bboxes