ML Inference Server

We build a simple image classification server using a pre-trained Densenet-121 model for Pytorch. The model is used for inference and is hosted on a backend Flask server built to handle HTTP requests, containerized in a Docker image.

• Project description: https://docs.google.com/document/d/1TxfuBqIY-Xo_NO9u9Kn5sdrAghRIFOvxL36FzVHoCU0/
• Team Members:
  • Dhruv Agarwal (dagarwal@umass.edu)
  • Shantam Shorewala (sshorewala@umass.edu)
  • Shubham Shetty (shubhamshett@umass.edu)

Test Script

The below command pulls the docker image from DockerHub, starts a container, and sends a classification request for data/dog.jpeg:

sh docker_run.sh 

Step-wise Instructions

• To download the pre-built image from DockerHub:

docker pull sshorewala/img-class-server

• To start a container from the pulled image:

docker run -p 5000:5000 sshorewala/img-class-server --name mlinf-team-7

• To run inference against the provided test image, execute (from project root):

curl -F "query=@data/images/zeppelin.jpg" http://localhost:5000/v1/densenet-inference/prediction

• To build the docker image locally:

docker build -t img-class-server .