Software
PSIAP-Video-Classifier Deployment
This package sets up a pair of Docker container services for classifying footage with a tensorflow model
The Services
- tf_server: A tensorflow-serving server that hosts convolutional neural networks. The example model classifies footage as in-vehicle or out-of-vehicle. The services can be accessed from localhost at localhost:8501, or from within the docker-compose network at tf_server. See the tf_server README for more details
- client: A python3 docker image that processes the videos contained within the subdirectory videos and generates a report in the subdirectory reports. It accomplishes this by using ffmpeg to extract keyframes, then submits those frames to tf_server for classification, and then generates a report based on the classifications. See the client README for more details