Amazon SageMaker

Prerequisites for consuming the service:

1. Access to Model Package, SageMaker and S3 storage bucket.
2. Open Trailer Container Images. (Refer to Sample Input linked below)
3. Execution Role for the SageMaker session.
4. Python Packages as listed in the Instructions Notebook linked below.

Input

Supported Content Type: 'application/json' (Image serialized to json as shown below in Python)

from PIL import Image
import json
import numpy as np

img = Image.open('images/sample1.jpg').convert(mode = 'RGB')
img = img.resize((300,300))

img = np.array(img).tolist()
img_json = json.dumps({'instances': [{'input_image': img}]})

// If required can be written to file (Also can be found in Sample link below)
with open('img.json', 'w') as f:
    f.write(img_json)

Output

Content Type: 'application/json'

Sample Output:

{"predictions": [[0.0353902504, 0.55054903, 0.414060771]]}

Invoking Endpoint

If you are using real time inferencing, please create the endpoint first.

Python

// Find detailed instructions in the Instructions Notebook lined below
predictor = sage.RealTimePredictor(endpoint='endpoint name', 
                                   content_type='application/json',
                                   sagemaker_session= sagemaker_session, )
prediction = predictor.predict(img_json)

AWS CLI Command

aws sagemaker-runtime invoke-endpoint --endpoint-name "endpoint-name" --body fileb://img.json --content-type application/json --accept application/json out.json

Notebook Instructions:

1. Download the Notebook from the link below onto a SageMaker Notebook Instance OR Install necessary packages on the desired compute resource.
2. Bring in the input images for classification onto the SageMaker Notebook Instance OR on the desired compute resource and follow the instructions in the Notebook.

Resources

Sample Input Sample Notebook

Mphasis DeepInsights
Usage Instructions Notebook
Sample Input Images