patrickvonplaten / AdvertisementDetection

AdvertisementDetection

In this university project, we manually labeled 1000 pictures sampled from german football videos und used the VGG-16 model architecture to classify the pictures to "showing advertisment panel" or "not showing advertisment panel".

How to run code

• install python3
• check createFramesFromVideo.sh - path of video file should correspond to to that in code
• install ffmpeg
• pip install opencv-python
• pip install tabulate
• install tensorflow
• pip install keras

1. create frames from video using createFramesFromVideo.sh
   • ./createFramesFromVideo.sh 0.005 -> 0.005 frames per second
2. label frames
3. run python advertisementDetection.py