Python implementation of the dynamic image technology discussed in 'Dynamic Image Networks for Action Recognition' by Bilen et al. Their paper and GitHub can be found as follows:
Clone the directory, and install the module and it's pre-requisites by running:
python setup.py install
import glob
import cv2
from dynamicimage import get_dynamic_image
def main():
frames = glob.glob('./example_frames/*.jpg')
frames = [cv2.imread(f) for f in frames]
dyn_image = get_dynamic_image(frames, normalized=True)
cv2.imshow('', dyn_image)
cv2.waitKey()
if __name__ == '__main__':
main()
Source Video: https://www.youtube.com/watch?v=fXMDubfvoQE