/face_rekognition_AWS

Setting up and calling AWS Face Rekognition API through Raspberry PI

Project Name: face_rekognition_AWS

Setting up and calling AWS Face Rekognition API through Raspberry PI

Outline

  1. Introduction
  2. Set up AWS on Raspberry PI
  3. Camera Set Up on Raspberry PI
  4. Upload Images to Amazon S3 Bucket
  5. Upload Faces to Collection
  6. Identify Faces from Collection, Determine Emotions and Display on Web browser
    • Identifying Emotions
    • Upload images to S3 bucket and create collection
    • Finding faces in the Database
    • Opening Web Browser using Python
  7. Putting it together

Introduction

Following repo will provide a guide on how to set up your Amazon AWS Rekognition, set up camera on Raspberry PI, identify emotions of the people in the picture and classify each face from known database.

Set up AWS on Raspberry PI

Installation is straighforward and pretty easy when you follow Installation Guide. Then, set up configuration details

Here are some of the applications you have to install before proceeding further: python3, python2, boto3, boto, opencv, PIL

Camera Set Up on Raspberry PI

  • Hardware

    Installing and attaching camera on Raspberry PI is straightforward, which can be found here

  • Software

    Python code camera.py will start camera on Raspberry PI through web browser. This python code already includes HTML code and some CSS codes

    Note You should have PIL (Python Image Library), boto3, picamera packages installed

    In order to run camera.py type in command prompt

    python3 camera.py

    Then, open preferred web browser and type

    http://0.0.0.0:8000/index.html

    Once you snap the picture on your web browser, it will create and save image currIm.jpg

Upload Images to S3 Bucket

  1. Create bucket on AWS S3. Then open upload_s3.py file in text editor

  2. Go to your Console on AWS and find AWS Access and Secret keys. Then, add Access and Secret keys to upload_s3.py file

    AWS_ACCESS_KEY = Your Access Key
AWS_ACCESS_SECRET_KEY = Your Secret Key

  3. From file upload_s3.py you can call function upload_main(fileName, bucket) by providing name of the file you want uploaded

Upload Faces to Collection

Note I found concept of collection a bit confusing at first. Essentially, every bucket can contain a collection. Moving forward, you will create bucket in S3 and upload your files into it. Bucket will contain physical copies of your files. Then, using those files in S3 bucket, you will create collection items, which can be characterized as virtually grouped items in S3

Open file collection_faces.py and do the following:

  • Replace variable value rootdir to a directory where all the faces are located. In this case, all the people you want to have in your dataset as known people
  • Replace bucket and collectionId with bucket and collection where you want to store your dataset of known people.
  • Helper functions you can (un)comment:
    • list_collection() will list all the collections in your Amazon AWS
    • list_items_collection(collectionId, bucket) takes 2 arguments (collection name and bucket name), and will return list of items in the collection & bucket specified.
  • Function get_bucket_names(bucket) takes 1 argument (bucket name) and returns a list of all the items in the bucket. This function will be used in the next step.
  • Function add_faces_collection(collectionId, photo_l, bucket) will take items from photo_l list and put them in collection and bucket you specified. Result will be a bucket & collection which will contain your known people dataset

Note The following code has been used from Amazon AWS Rekognition Documentation

Identify Faces from Collection, Determine Emotions and Display on Web browser

Note: make sure to install opencv pip3 install opencv-python

When you run the following code in your terminal python test_collection.py, the code will do the following:

  1. It will wait until currIm.jpg file appears in the folder
  2. Then it will run function recognition_main() which identifies all the new faces, and compares each new face with faces in your collection
    • In other words, suppose you created collection myCollection which contains some # of faces. Suppose file currIm.jpg already exists. After recognition_main() function is called all the faces identified in the currIm.jpg will be added to myCollection and compared with all the faces in the collection.
    • In Amazon AWS (as far as I know) in order to recognize and identify many faces in the image, you have to upload all the images to the same collection. Example: Suppose your collection myCollection contains 100 people. Also, suppose new picture you uploaded is a group picture that has 5 faces, and you want to see which of these 5 people are in myCollection. In order to do that, you have to upload all the faces the same collection
  3. Once all the faces are identified, it will run emotion_main() function, which is located in demo.py python file.
    • In the demo.py file, replace bucket variable inside emotion_main() function
    • The result will be, new file currIm_emotion.jpg with all the emotions identified for each face in the currIm.jpg image.
  4. Once both emotion_main() and recognition_main() executed, the result will be projected on web browser of your choosing automatically.

Note The following code has been used from Amazon AWS Rekognition Documentation

Putting it all together

  1. Open terminal and type python3 camera.py
  2. Open web browser and type http://0.0.0.0:8000/index.html
  3. Open new terminal window, or new tab and type python test_collection.py
    • You can also run both camera.py and test_collection.py in the same terminal window. In that case, type
    $ python camera.py &
$ python test_collection.py
  4. Result webpage with 3 images: original image (currIm.jpg), identified faces image, and emotion + identified faces