aws_auto_checkin_app.py

What is this?

Face recognition using Amazon Rekognition, see SearchFacesByImage and Auto Check-In App for more info.

Subscribing Topic

• image/compressed (sensor_msgs/CompressedImage)

  Input compressed image. (only when ~image_transport is compressed.)

  This topic name is resolved from image.

• image (sensor_msgs/Image)

  Input raw image. (only when ~image_transport is raw.)

• face_roi (opencv_apps/FaceArrayStamped)

  Rectangles on the face of input image. Use ROI value.

msg.faces[].face.x      : X coordinates of the center of the face image in the ~image input
msg.faces[].face.y      : Y coordinates of the center of the face image in the ~image input
msg.faces[].face.width  : Width of the face image
msg.faces[].face.height : Height of the face iamge

Publishing Topic

• face_name (opencv_apps/FaceArrayStamped)

  Publish recognized face name as well as face image. The face.{x,y,width,height} corresponds to input face_roi, that means x, y is the center of face rectangle.

• ~output/rects (jsk_recognition_msgs/RectArray)

  Rectangles of matched faces.

• ~output/class (jsk_recognition_msgs/ClassificationResult)

  Detected face class labels and probabilities.

Parameters

• ~use_window (Bool, default: False)

  Show input image on the window, if it is true.

• ~approximate_sync (Bool, default: True)

  Approximately synchronize inputs if it’s true.

• ~queue_size (Int, default: 100)

  How many messages you allow about the subscriber to keep in the queue. This should be big when there is much difference about delay between two topics.

• ~slop (Float, default: 1.0)

  Maximum allowed time for approximate synchronization in [sec].

  This parameter is enabled only when ~approximate_sync is True.

• ~classifier_name (String, default: rospy.get_name())

  Classifier name written to classifier field of ~output/class.

• ~image_transport (String, default: compressed)

  Set raw to subscribe raw image.

• ~env_path (String, default: env.json)

  Json file for environment variables.

  See DynamoDB for how to obtain DynamoDB Table name and Rekognition for CollectionId.

{
  "Region": "%%REGION%%",
  "DynamodbTable": "%%DYNAMODB_TABLE%%",
  "CollectionId": "%%COLLECTION_ID%%",
  "FaceAreaThreshold": 1e4,
  "FaceSimilarityThreshold": 90,
  "MaxFaces": 1
}

• ~aws_credentials_path (String, Default: aws.json)

  Json file for aws access key and secret key to run AWS Rekognition. See Set up an AWS account and create an IAM user for how to obtain keys.

{
    "aws_access_key_id" : "####################",
    "aws_secret_access_key" : "********************"
}

Sample

roslaunch jsk_perception sample_aws_auto_checkin_app.launch use_aws_face_detection:=true

If use_aws_face_detection is true, AWS face detection API is used. Otherwise OpenCV face detection is used. AWS face detection is more accurate.

For JSK user, Download env.json file and aws.json from Gdrive and put these under /tmp directory to run sample code.

To add new people to face database, add face image file to Amazon S3, auto-check-in-gapp-register… buckets

For more info about managing the face collection, see Searching faces in a collection.