atakanemree/image-caption-api repository overview

⁠🖼️ Image Captioning API

Production-ready Image Captioning Service powered by PyTorch and FastAPI. Generate human-like captions for images with a single API call.

⁠🚀 Quick Start

The easiest way to run the API is using Docker:

docker run -p 8000:8000 atakanemree/image-caption-api:latest

Once running, access the API documentation at: http://localhost:8000/docs⁠

⁠✨ Features

• 🧠 Advanced Architecture: ResNet50 Encoder + LSTM Decoder with Attention
• ⚡ High Performance: Optimized for speed (~50ms inference on CPU)
• 🐳 Docker Ready: Pre-built images for CPU and GPU
• 🔍 Beam Search: High-quality caption generation
• 🛡️ Robust: Input validation, error handling, and health checks

⁠🎯 Usage

⁠1. Health Check

curl http://localhost:8000/health

⁠2. Generate Caption

curl -X POST \
  http://localhost:8000/predict \
  -H 'Content-Type: multipart/form-data' \
  -F '[email protected]'

Response:

{
    "predicted_caption": "a dog running on the grass",
    "confidence": 0.85,
    "timestamp": "2023-11-29T12:00:00"
}

⁠3. Batch Prediction

curl -X POST \
  http://localhost:8000/predict/batch \
  -H 'Content-Type: multipart/form-data' \
  -F '[email protected]' \
  -F '[email protected]'

⁠🐳 Docker Variants

⁠Running with GPU

docker run --gpus all -p 8000:8000 atakanemree/image-caption-api:latest-gpu

⁠🔧 Configuration

You can configure the container using environment variables:

docker run -p 8000:8000 \
  -e API_WORKERS=4 \
  -e LOG_LEVEL=INFO \
  atakanemree/image-caption-api:latest

⁠🏗️ Architecture

The system consists of a ResNet50 CNN for image feature extraction and an LSTM RNN for text generation. The API is built with FastAPI for high performance and easy integration.

For more details, source code, and training instructions, visit the GitHub Repository⁠.