This repository documents a learning project that explores multiple AWS AI services to analyze images, extract text, and generate summaries.
The project evolves through three stages:
- OCR using Amazon Textract
- Text summarization and sentiment analysis using Amazon Comprehend
- Image understanding using Amazon Bedrock (Claude 3)
The goal of this project is to understand how different AWS AI services can be combined to process images and extract meaningful insights. The workflow progresses from traditional OCR-based extraction to foundation-model-based image understanding.
- Input: JPG / PNG images
- Service used: Amazon Textract
- Output: Extracted text saved as
.txtfiles
- OCR: Amazon Textract
- Text analysis: Amazon Comprehend
- Output:
- Extracted text
- Key phrases
- Sentiment classification
- Input: Image sent directly to a foundation model
- Model used: Anthropic Claude 3 (via Bedrock)
- Output: Natural language explanation of image content
The following AWS services were enabled using the AWS Management Console for this learning project:
| Service | API / Model | IAM Policy | Notes |
|---|---|---|---|
| Amazon Textract | DetectDocumentText | FullAccess | For learning purposes |
| Amazon Comprehend | DetectKeyPhrases, DetectSentiment | FullAccess | Language: English |
| Amazon Bedrock | Claude 3 Haiku via Bedrock | FullAccess | Requests made using bedrock-runtime |
Notes:
- Full access policies were used to simplify experimentation.
- For production or sensitive projects, always follow the principle of least privilege.
- No credentials, account IDs, or sensitive information are included in this repository.
Below are example outputs generated by the scripts.
Screenshots are included for demonstration only. No sensitive or billable information is shown.
(Screenshots will be added soon.)
This project is based on learning material and tutorials exploring Amazon Textract, Comprehend, and Bedrock usage with Python. This repository is intended for educational purposes.