This solution provides an example of how to create an end-to-end solution to inference images of food/dishes and create a recipe.
You can access a deployed instance of this solution at https://recipe-generation.com - (For simplicity, the demo has a pre-defined set of images for the user to choose instead of allowing uploads. The the web portal code in this repo allows users to take pictures from their mobile phones)
The infrastructure is based on:
- Amazon API Gateway - Entry point to send images as base64 and return recipes
- Amazon Lambda Function - Processes the image invoking the BLIP2 and generates a recipe with Claude instant
- Amazon Cognito - Provides authentication for the API
- Amazon Sagemaker - Used to host the BLIP2 model
- Amazon Bedrock - Used for inference requests to Claude instant
Most of the deployment of the infrastructure is automated using the Serverless Application Model. For detail instructions, follows the steps here
Example single page application built in Vue JS - it is integrated with Cognito to provide a seamlessly experience, allowing users to sign up and sign in.
Detail instructions can be found here
Please open an issue for any questions or issues