extract_yolo_v3_retraining_data_from_video

Decompose video into individual frames with bounding box output formatted for Yolo v3 retraining.

TODO: ### DOI

Use object tracking software (OpenCV2) to generate yolov3 suitable training images from videos for new object types

• Obtain videos of the objects of interest
• test each video
  • Run python object_track_extract_yolov3.py test <video_path>
  • The first video frame is loaded
  • Draw a bounding box around the object of interest
  • Press the spacebar and verify that the bounding box properly follows the object without errors
    • If errors occur, change the tracker_types in the script to another type and try again
• capture data from each video
  • Run python object_track_extract_yolov3.py capture <video_path> <output_path>
  • Follow the same process as above.
  • With the 'capture' argument, data will be output to the <output path>

Environment Requirements

• Python 3.7
• OpenCV
  • Install cv2 in python3 via pip install opencv-contrib-python
  • The "contrib" installation expands the install to include all the object tracking packages

References

Object Tracking Video code modified from Satya Mallick's article on LearnOpenCV here: https://www.learnopencv.com/object-tracking-using-opencv-cpp-python/