Yizhang-Li / FSDL_ANOMALIB

Deploy Anomalib to AWS (Model)

Setup

Make sure you have conda installed. Then, install the requirements with bash install_requirements.sh.

Prepare MVTEC Data

python datasets/prepare_data.py

Download pre-trained Padim models

Unzip Padim models (including meta-data) to the folder results/.

You need to put the following files in the folder results/:

• results/padim/mvtec/bottle/openvino/model.bin
• results/padim/mvtec/bottle/openvino/meta_data.json

Gradio Inference (Local)

python tools/inference/gradio_inference.py --config tools/config/padim/config.yaml --weights results/padim/mvtec/bottle/openvino/model.bin --meta_data results/padim/mvtec/bottle/openvino/meta_data.json

Deployment

1. Create/setup a AWS account
2. Register an image repo on AWS ECR.
3. Upload your Docker image to AWS Elastic Container Registry (ECR)
   • docker login
   • docker build -t
   • docker tag
   • docker push
4. Create your AWS Lambda to run the ECR image
   • Create the Lambda function with pushed image
   • Config max memory usage and timeout
   • Create a test
   • Add a triger
5. Run the frontend in a seperate ec2 instance.

Demo

Demo