Canny Edge Detector Prototype
This is a microservice prototype, which runs the Canny Edge Detection algorithm on images uploaded to an REST endpoint. This was written in three hours as part of an interview assignment so expect bugs and errors.
Run
To run this microservice in debug mode, execute docker-compose up (or docker-compose up -d to run in the background). The default port is 80 and can be adjusted in the docker-compose.override.yml.
Use the service with any web request tool like Postman or cUrl.
Further Improvements
- Setup CI lint
- Use external uWSGI and Nginx
- Provide Helm chart
- Use numpy on the whole arrays instead of just the single values inside for-loops
- Perform the calculations in another thread (using OpenCV would also increase preformance significantly)
- Rearchitect the application
- Put the image into a queue and return transaction ID instead of the image
- Provide a different endpoint, which, given the transaction ID, returns the processed image (or an error if its not yet finished)
- Or build it cloud-native with AWS/Azure Queues and Lamdbas/Cloud Functions
Examples
Query (cURL)
curl --location --request POST 'localhost/detect' \
--form 'file=@"./input.png"' \
--output "output.png"Result
