PSIAP-Video-Classifier-Deployment

This sets up a pair of Docker container services for classifying footage with a tensorflow model

The two services are:

• tf_server: A tensorflow-serving server which 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 which 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

Windows Set-Up

0. Create a Docker Hub Account
1. Install Docker Desktop for Windows
2. Add Docker to Windows Path, should be something like C:\Program Files\Docker\Docker\resources\bin
3. Configure Docker to use proxy, if needed (note look for config.json in \Users\USERXXX.docker)
4. Copy video files to directory: \deploy\videos
5. Open Windows PowerShell
6. Navigate to deploy directory
7. Log in to the Docker registry, docker login
8. Run Docker Compose to build service, docker-compose build
9. docker-compose up
10. docker-compose up -d

*nix instructions

0. Install Docker for your platform, e.g. ubuntu
1. Configure proxy if necessary
2. Navigate into this directory in terminal
3. Copy video files to directory ./videos
4. docker-compose build
5. docker-compose up

Authors

• Jeffrey Liu (MITLL)
• Chris Budny (MITLL)
• Andrew Weinert (MITLL)

Acknowledgments

• Gabriela Barrera (MITLL)
• Dieter Schuldt (MITLL)
• Steven Talpas (NJOHSP)
• William Drew (NJOHSP)

Disclaimer

This work was performed under the following financial assistance award 70NANB17Hl69 from U.S. Department of Commerce, National Institute of Standards and Technology.