lucaspellucci / train_amazon_rekognition

Train Amazon Rekognition

Train Amazon Rekognition with Face Images

Simple example of a Docker container that creates a collection on Amazon Rekognition (https://aws.amazon.com/rekognition/) and adds faces from image files to it.

Build

make (build|rebuild)

• This will create a Container Image to be used.

Setting credentials

Set the AWS credentials that will passed to the container in runtime.

export AWS_ACCESS_KEY_ID=AKIA***

export AWS_SECRET_ACCESS_KEY=wJalrXU***

export AWS_DEFAULT_REGION=ap-southeast-2

• Alternatively, you can print the credentials from the AWS cli and copy the export commands:

make print-creds #for default profile or make print-creds PROFILE=my_profile

• Copy this to your shell to set the credentials as Environment Variables

Usage

• Create a /img folder with the images you want to train, the suggestion is to have the filenames as the person name, like joao-silva001.jpg, joao-silva002.jpg, jose-carlos001.jpg, and so on...

make run COLLECTION="-c collection-name"

• will add all images in /img to the collection called "collection-name" and will use the file name (without the extension) as a ExternalID.

License

This project is licensed under the Apache 2.0 License.