metadata

title: Object Detection Lambda
emoji: 🌖
colorFrom: purple
colorTo: green
sdk: gradio
sdk_version: 5.5.0
app_file: app.py
pinned: false
short_description: Object detection Lambda

Object detection via AWS Lambda

Aim: AI-driven object detection task

Architecture:

Front-end: user interface via Gradio library
Back-end: use of AWS Lambda function to run deployed ML model

You can try out our deployed Hugging Face Space!

Table of contents:

Local development
AWS deployment
Hugging Face deployment

1. Local development

1.1. Build and run the Docker container

Step 1 - Building the docker image

bash

docker build -t object-detection-lambda .

Step 2 - Running the docker container locally

bash

docker run --name object-detection-lambda-cont -p 8080:8080 object-detection-lambda

1.2. Execution via user interface

Use of Gradio library for web interface

Note: The environment variable AWS_API should point to the local container

export AWS_API=http://localhost:8080

Command line for execution:

python3 app.py

The Gradio web application should now be accessible at http://localhost:7860

1.3. Execution via command line:

Example of a prediction request

bash

encoded_image=$(base64 -i ./tests/data/boats.jpg)

curl -X POST "http://localhost:8080/2015-03-31/functions/function/invocations"
-H "Content-Type: application/json"
-d '{"body": "'"$encoded_image"'", "isBase64Encoded": true, "model":"yolos-small"}'

python

python3 inference_api.py
--api http://localhost:8080/2015-03-31/functions/function/invocations
--file ./tests/data/boats.jpg
--model yolos-small

2. Deployment to AWS

2.1. Pushing the docker container to AWS ECR

Steps:

Create new ECR Repository via aws console

Example: object-detection-lambda

Optional for aws cli configuration (to run above commands):

aws configure
Authenticate Docker client to the Amazon ECR registry

aws ecr get-login-password --region | docker login --username AWS --password-stdin .dkr.ecr..amazonaws.com
Tag local docker image with the Amazon ECR registry and repository

docker tag object-detection-lambda:latest .dkr.ecr..amazonaws.com/object-detection-lambda:latest
Push docker image to ECR

docker push .dkr.ecr..amazonaws.com/object-detection-lambda:latest

Link to AWS ECR Documention

2.2. Creating and testing a Lambda function

Steps:

Create function from container image

Example name: object-detection

Notes: the API endpoint will use the lambda_function.py file and lambda_hander function
Test the lambda via the AWS console

Advanced notes:

Steps to update the Lambda function with latest container via aws cli:

aws lambda update-function-code --function-name object-detection --image-uri .dkr.ecr..amazonaws.com/object-detection-lambda:latest

2.3. Creating a REST API via API Gateway

Steps:

Create a new Rest API (e.g. object-detection-api)
Add a new resource to the API (e.g. /detect)
Add a POST method to the resource
Integrate the Lambda function to the API
- Notes: currently using proxy integration option unchecked
Deploy API with a specific stage (e.g. dev stage)

Example AWS API Endpoint: https://<api_id>.execute-api.<aws_region>.amazonaws.com/dev/detect

2.4. Execution for deployed model

Example of a prediction request

bash

encoded_image=$(base64 -i ./tests/data/boats.jpg)

curl -X POST "https://.execute-api..amazonaws.com/dev/detect"
-H "Content-Type: application/json"
-d '{"body": "'"$encoded_image"'", "isBase64Encoded": true, "model":"yolos-small"}'

python

python3 inference_api.py
--api https://.execute-api..amazonaws.com/dev/detect
--file ./tests/data/boats.jpg
--model yolos-small

3. Deployment to Hugging Face

This web application is available on Hugging Face

Hugging Face space URL: https://huggingface.co./spaces/cvachet/object_detection_lambda

Note: This space uses the ML model deployed on AWS Lambda