folkien/yaya

YAYA - Yet annother YOLO annoter for images (in QT5). Support yolo format, image modifications, labeling and detecting with previously trained detector.

YAYA - Yet Another YOLO Annoter

YAYA - Yet Another YOLO Annoter with QT5 widgets gui, and ...

• Rewritten in python,
• Checks for errors of overriding boxes,
• Displays image properties size, hue, saturation, brightness,
• Displays annotations properties, average size, class numbers,
• Uses given YOLOv4 detectors to detect every file and store detections!
• Calculate metrics TP,TN,FP,FN,Precision,Recall for every photo!
• Auto-annotation with YOLOv4 detectors feature added - use yolo to detect and describe annotations of your image,
• Manual Yolo detection by presing 'd' - to check YOLO with original data,
• You can use standard YOLOv4 (MSCOCO) or your custom YOLOv4

Requirements

pip install -r requirements

Install YOLOv4 darknet library libdarknet.so in your operating system (https://github.com/AlexeyAB/darknet) for usage of custom YOLOv4 detectors.

How to add custom YOLOv4 detector?

1. Inside directory ObjectDetectors/ create your detector directory (for example yolov4custom).
2. Copy all YOLOv4 detector files : yolo.cfg, yolo.data, yolo.names, yolo.weights (names should be identicall)
3. Got it! Now you can use this detector!

Found detectors list is also shown at the program start, example :

python ./yolo-annotate.py -i input/
DEBUG:root:Logging enabled!
/usr/local/lib/libdarknet.so
INFO:root:(Found detector) 0 - /home/spasz/python/aisp-tools/yaya/ObjectDetectors/yolov4custom/yolo.

How to start?

To load all test images from input directory and start application, you can use command

./yolo-annotate.py -i input/

Key codes

LPM - create annotation
d - run detector
r - remove annotation
c - clear all annotations
s - save all (if errors not exists)
arrow -> or . - next image
arrow <- or , - previous image

Command line

usage: yolo-annotate.py [-h] -i INPUT [-c CONFIG] [-on] [-yc] [-v]

optional arguments:
  -h, --help       show this help message and exit
  -i INPUT, --input INPUT
             Input path
  -c CONFIG, --config CONFIG
             Config path
  -on, --onlyNewFiles  Process only files without detections file.
  -yc, --yoloCustom   Use custom YOLO.
  -v, --verbose     Show verbose finded and processed data

---

[!DEPRECATED!] yolo-annotate - old OpenCV version.

[!DEPRECATED! - Use release/v0.6-OpenCV for old OpenCV version] Yet Another yolo annotation program. Yolo_mark clone with openCV gui, but ...

• Rewritten in python,
• Checks for errors of overriding boxes,
• Auto-annotation feature added - use yolo to detect and describe annotations of your image,
• Manual Yolo detection by presing 'd' - to check YOLO with original data,
• You can use standard YOLOv4 (MSCOCO) or your custom YOLOv4