A cloud-native computer vision pipeline that automatically processes images, detects multi-class objects, and dynamically renders spatial bounding boxes using deep learning.
This project leverages AWS deep learning models to perform automated object detection and spatial analysis on uploaded images. By coupling cloud storage with serverless vision APIs, it extracts high-confidence labels and maps bounding box coordinates natively onto local image renders.
- Smart Surveillance: Automated anomaly and object detection in real-time streams.
- Retail Inventory: Automated item counting and category tagging for digital shelves.
- Accessibility Tech: Generating descriptive spatial metadata for visually impaired users.
[ Raw Image ] ──> ( Amazon S3 Bucket ) ──> [ Python Script + Boto3 ] │ ▼ [ Matplotlib Render ] <── [ Bounding Box Coordinates ] <── ( Amazon Rekognition )
- Ingestion: Raw images are staged securely in an Amazon S3 repository.
- Analysis: A lightweight Python runtime dispatches the image payloads to Amazon Rekognition via the
detect_labelsAPI. - Authorization: Granular access boundaries are enforced via AWS IAM service roles.
- Visualization: The engine parses spatial coordinate geometry from the API JSON response, leveraging Matplotlib to map and overlay high-contrast bounding boxes.
A computer vision project that automatically detects objects in images and draws bounding boxes with confidence scores — powered by AWS Rekognition and Python.
Upload any image and the script will:
- Detect all objects in the image using AWS deep learning AI
- Label each object with its name and confidence percentage
- Draw colored bounding boxes around every detected object
- Save the annotated image to the
output/folder
| Tool | Purpose |
|---|---|
| Python 3 | Main programming language |
| AWS S3 | Cloud storage for uploaded images |
| AWS Rekognition | AI image analysis (object detection) |
| boto3 | Python library to talk to AWS |
| matplotlib | Drawing bounding boxes on images |
| Pillow | Reading and processing image files |
Image Labels Generator/
├── images/ ← Put your input images here
├── output/ ← Annotated images saved here (auto-created)
├── screenshots/ ← Demo images for this README
├── detect_labels.py ← Main script
└── requirements.txt ← Python dependencies
- Python 3.8+
- An AWS account (free tier is enough)
- AWS CLI installed and configured
git clone https://github.com/YOUR-USERNAME/image-labels-generator.git
cd image-labels-generator# Windows
python -m venv venv
.\venv\Scripts\Activate.ps1
# Mac/Linux
python -m venv venv
source venv/bin/activatepip install -r requirements.txtaws configure
# Enter your Access Key ID, Secret Access Key, region (us-east-1), output (json)Go to AWS Console → S3 → Create bucket. Note your bucket name.
Open detect_labels.py and update line 10:
BUCKET_NAME = "your-bucket-name-here"Add any .jpg image to the images/ folder, then run:
python detect_labels.py images/sample.jpgExample output:
Uploading sample.jpg to S3...
Upload complete.
Analyzing image with Rekognition...
Detected 7 labels:
- Dog (100.0%)
- Animal (100.0%)
- Mammal (100.0%)
- Pet (100.0%)
- Puppy (100.0%)
- Hound (80.8%)
Output saved to: output\labeled_sample.jpg
Open the result:
# Windows
start output\labeled_sample.jpgYour Image → Upload to S3 → Rekognition API → Parse Labels → Draw Boxes → Save Output
- Upload — Image is uploaded to your S3 bucket
- Detect — Rekognition analyzes the image and returns detected labels with bounding box coordinates
- Draw — matplotlib draws colored rectangles around each object using the coordinates
- Save — Annotated image is saved to
output/
This project runs entirely within the AWS Free Tier:
| Service | Free Allowance |
|---|---|
| Amazon S3 | 5 GB storage, 20k GET requests/month |
| Amazon Rekognition | 5,000 images/month (first 12 months) |
- Smart surveillance and security systems
- Automated retail inventory management
- Accessibility tools for visually impaired users
- Automatic photo organization and tagging
- Content moderation for user-uploaded images