Process video frame by frame by detecting objects using Keras.. For illustration, the scripts are used to identify animal species in wild-life videos..
[1] To process a single image input,
Usage: python3 keras-vgg16.py IMAGE
Sample images are provided in 'sample-images' folder
Top 5 most relevant tags are printed directly on stdout..
[2] To process video,
[A] If you prefer to run python script via command-line:
Usage: python3 process-video.py VIDEO
Sample video is provided for testing in 'sample-videos' folder
Output of process-video.py script is stored in 'result/output.mp4' and shows
the most relevant tags for each frame
[B] Jupyter Notebook: video-tagger.ipynb
Follow the step-by-step instructions for processing videos using object detection..
Provides the option to either download an online video using youtube-dl library, or
run video-tagger on local video file