AWS Rekognition + OpenCV

Here is how you can use OpenCV with AWS rekognition

1. Enable your configured aws account on rekognition and S3
2. Create a rekognition collection this will be the collection of faces the camera stream is matched against > #aws rekognition create-collection –collection-id “faces” –region us-east-1
3. List collections to make sure that the collection was created #aws rekognition list-collections –region us-east-1
4. Create an S3 bucket and upload the photo with the face you’d like your camera to rekognize
5. Index the faces in your S3 Bucket
   aws rekognition index-faces \
   –image ‘{“S3Object”:{“Bucket”:”BucketName“,”Name”:”FileName.png“}}’ \
   –collection-id “faces” \
   –region us-east-1
6. The following python code should work

import boto3

import cv2

from PIL import Image

import io

frame_skip = 12 # analyze every 100 frames to cut down on Rekognition API calls

threshold=80

vidcap = cv2.VideoCapture(0)

cur_frame = 0

success = True

BUCKET = “BUCKETNAME” ### This is the bucket where the face your trying to match lives

COLLECTION = “COLLECTIONNAME” #### This is the collection of faces that you have already taught the ai to store

def search_faces_by_image(bin_img, key, collection_id, threshold=80, region=”us-east-1″):

rekognition = boto3.client(“rekognition”, region)

try:

response = rekognition.search_faces_by_image(

Image={‘Bytes’: bin_img},

CollectionId=collection_id,

FaceMatchThreshold=threshold,

)

return response[‘FaceMatches’]

except:

print(“no faces found”)

return

while success:

success, frame = vidcap.read() # get next frame from video

if cur_frame % frame_skip ==0: # only analyze every n frames

print(‘frame: {}’.format(cur_frame))

pil_img = Image.fromarray(frame) # convert opencv frame (with type()==numpy) into PIL Image

stream = io.BytesIO()

pil_img.save(stream, format=’JPEG’) # convert PIL Image to Bytes

bin_img = stream.getvalue()

face = search_faces_by_image(bin_img, KEY, COLLECTION)

print(face)

cur_frame += 1