Yolo2Pascal-annotation-conversion

Yolo2Pascal-annotation-conversion copied to clipboard

Yolo2Pascal-Annotation-Conversion

Perform conversion between YOLO annotation format and PASCAL VOC format and vice versa. Code based on LabelImage repo.

Usage:

• Images must be in .jpg/.png format.
• YOLO to PASCAL: The script will search for all .txt files in the folder and perform the conversion. Converted PASCAL annotation files have a .xml extension, reside in the same folder and share the same name with their corresponding images.
• PASCAL to YOLO: The script will search for all .xml files in the folder and perform the conversion. The script requires a file class.txt describing all classes (this file is generated autonomously if using the LabelImage tool above for annotating images). Converted YOLO annotation files have a .txt extension, reside in the same folder and share the same name with their corresponding images.

Demo:

• YOLO to PASCAL:
  • python3 yolo2pascal/yolo2voc.py demo/yolo2pascal
  • After this, 2 files cat.xml and dog.xml should appear in the demo/yolo2pascal folder. They are label files in PASCAL format. You can use LabelImage to load the directory using PASCAL format to check.
• PASCAL to YOLO:
  • python3 pascal2yolo/voc2yolo.py demo/pascal2yolo
  • After this, 2 files cat.txt and dog.txt should appear in the demo/pascal2yolo folder. They are label files in YOLO format. You can use LabelImage to load the directory using YOLO format to check.
• Example of class.txt file. They are just labels listed in an alphabetical order:
  • cat
  • dog

TODO:

• [ ] Remove Qt dependency
• [x] Add support for different image formats (thanks guysoft)