AWS Lambda 将图像文件发送到 Amazon Sagemaker

Question

I have the next problem, I try send image file in base64 to Lambda function(write in Python), for invoke Sagemaker Endpoint, this is my Lambda Function:

import os
import io
import boto3
import json
import base64

# grab environment variables
ENDPOINT_NAME = os.environ['ENDPOINT_NAME']
runtime= boto3.client('runtime.sagemaker')

def lambda_handler(event, context):
    print("Received event: " + json.dumps(event, indent=2))
    
    data = json.loads(json.dumps(event))
    payload = data['foto']
    img = base64.b64decode(payload)
    #body = json.dumps({"instances": img})

    try:
        response = runtime.invoke_endpoint(EndpointName=ENDPOINT_NAME,
                                            Body=img)
        print(response)
    except Exception as e:
        print("Inference Error:")
        print(e)

    return { "img":"ok" }

And sagemaker endpoint invoke in Jupyter works fine with this code,

from tensorflow.keras.preprocessing.image import ImageDataGenerator

test_datagen = ImageDataGenerator(rescale=1./255)
test_generator = test_datagen.flow_from_directory('./data/val', target_size=(224,224))

instance = test_generator[1][0]
print(instance.shape)
array = instance.reshape((1,) + instance.shape)
payload = { 'instances': array.tolist() }

resp = tf_predictor.predict(payload)['predictions']
print(resp)

The error in Lambda says that the data type is unknown, and CloudWatch error in dimensions.

I think, that the error is in Body of invoke_endpoint method, but i not found how convert bytes type to list with image data. I need custom Lambda Function with numpy, or other library? or isn't necesary

Answer 1

How are you invoking the Lambda function? If you are using API Gateway to pass the image data to the SageMaker Runtime endpoint, you may need to set the contentHandling properties on the integration to handle binary encoding. See here:

https://docs.aws.amazon.com/apigateway/latest/developerguide/api-gateway-payload-encodings.html