Gemini Image Analyzer

An AI-powered image analysis service built with FastAPI and Google Gemini Vision API.

The application accepts image uploads or Base64-encoded images, validates and preprocesses them, sends them to the Gemini Vision model for analysis, and returns a clean, structured JSON response suitable for frontend applications and downstream systems.

---

Features

• AI-powered image analysis using Google Gemini Vision API
• Upload images (JPEG, PNG, GIF, BMP)
• Analyze images from Base64 strings
• Multiple analysis modes
  • General Analysis
  • Image Classification
  • Text Extraction
  • Jewelry Product Analysis
• Image validation and preprocessing
• Automatic image resizing
• Structured JSON responses
• API Key authentication
• Docker support
• Unit tests
• Interactive API documentation using Swagger

---

Tech Stack

Backend

• Python
• FastAPI
• Pydantic
• Uvicorn

AI

• Google Gemini Vision API

Image Processing

• Pillow (PIL)

Security

• API Key Authentication

Deployment

• Docker

Testing

• Pytest

---

Project Architecture

                Client
                   │
        Upload Image / Base64
                   │
                   ▼
          FastAPI REST API
                   │
         Request Validation
                   │
                   ▼
      Image Preprocessing Layer
      --------------------------
      • Validate image format
      • Validate size
      • Resize large images
                   │
                   ▼
         Gemini Client Layer
                   │
        Google Gemini Vision API
                   │
                   ▼
           AI Response
                   │
                   ▼
        Response Normalizer
                   │
                   ▼
        Structured JSON Output

---

API Endpoints

Analyze Uploaded Image

POST /analyze

Consumes:

• Multipart Form Data
• UploadFile

Parameters

Parameter	Type	Required
file	UploadFile	Yes
analysis_type	Form	No

Example

POST /analyze

Returns

{
  "status": "success",
  "analysis": {
    "description": "...",
    "labels": [],
    "confidence": 0.98
  },
  "errors": []
}

---

Analyze Base64 Image

POST /analyze-base64

Supports

• Base64 Images
• Data URI
• Local File Path

---

Generate API Key

POST /admin/generate-key

Creates API keys for authenticated users.

---

Image Processing Pipeline

1. Receive uploaded image
2. Validate file size
3. Validate image format
4. Normalize image
5. Resize if required
6. Send image to Gemini API
7. Receive AI response
8. Normalize output
9. Return structured JSON

---

Supported Analysis Types

General

General image understanding.

Classification

Identify objects and categories.

Extraction

Extract text and information.

Jewelry

Specialized prompt that extracts

• Product Name
• Jewelry Type
• Metal
• Stone
• Color
• SEO Title
• SEO Description
• Marketing Text
• Confidence Score

---

Running Locally

Clone

git clone https://github.com/dipayan3203/Image-Analyzer.git

Install

pip install -r requirements.txt

Environment Variables

Create a .env

GEMINI_API_KEY=YOUR_KEY
GEMINI_MODEL=gemini-2.5-flash

Run

uvicorn app.main:app --reload

---

API Documentation

Swagger UI

http://localhost:8000/docs

---

Testing

pytest

---

Future Improvements

• Batch image analysis
• Async background processing
• Rate limiting
• Response caching
• Database persistence
• User authentication (JWT)
• Image history
• Multi-model AI support

---

Folder Structure

app/
│
├── main.py
├── gemini_client.py
├── middleware.py
├── normalizer.py
├── preprocessor.py
├── schemas.py
├── database.py
│
├── static/
│
tests/
│
Dockerfile
requirements.txt

---

Key Learnings

• FastAPI REST API development
• Google Gemini Vision API integration
• Image preprocessing using Pillow
• API authentication
• Response normalization
• Dockerized deployment
• Error handling
• AI application development

---

Author

Dipayan Mahato

Python Backend Developer