grycap/darknet repository overview

⁠Running Darknet on AWS Lambda

You can run Darknet⁠ in AWS Lambda via SCAR⁠ using the grycap/darknet⁠ Docker image, based on the bitnami/minideb:jessie⁠ one.

Darknet⁠ is an open source neural network framework written in C and CUDA. For the example we will be using the the library 'You only look once' Yolo⁠ which is is a state-of-the-art, real-time object detection system

Since we are using Darknet on the CPU it takes around 6-12 seconds per image, using the GPU version would be much faster.

⁠Usage in AWS Lambda via SCAR

You can run a container out of this image on AWS Lambda via SCAR⁠ using the following procedure:

1. Create the Lambda function

scar init -s yolo-sample-object-detection.sh -es s3-bucket -m 1024 grycap/darknet

1. Launch the Lambda function uploading a file to the /input folder of the specified S3 bucket

Take into consideration than the first invocation will take considerably longer than the subsequent ones, where the container will be cached.

1. In the output folder of the S3 bucket you will see the output of the yolo system with the name of the image that you uploaded