You can deploy this model and start earning money today!

Kishalay6/
ImageCaptionGenerator

Views

No views yet

🖼️ ImageCaption — AI Image Captioning

An AI-powered image captioning system that generates human-readable descriptions for uploaded images using a VGG16 + LSTM deep learning architecture.

Mini Project — Kalinga Institute of Industrial Technology (KIIT), Bhubaneswar

✨ Features

📷 Drag & drop image upload (JPG, JPEG, PNG)
🧠 Deep learning caption generation using VGG16 feature extraction + LSTM decoder
⚡ Flask REST API backend
⚛️ React frontend with a modern dark UI
🎨 Animated glassmorphism design with gradient accents

🛠️ Tech Stack

Layer	Technology
Frontend	React 18 (CDN), CSS3
Backend	Flask, Flask-CORS
ML Model	TensorFlow / Keras
Feature Extraction	VGG16 (ImageNet)
Caption Decoder	LSTM (Greedy decode)
Dataset	Flickr8k

🚀 Getting Started

Prerequisites

Python 3.12 (required for TensorFlow compatibility)
pip package manager

Installation

bash

# Clone the repo
git clone https://github.com/OmPrakashMishra1/imageCaption.git
cd imageCaption

# Install dependencies
pip install -r backend/requirements.txt

Running the App

bash

cd backend
py -3.12 app.py

Open http://localhost:5000 in your browser — the React frontend is served directly by Flask.

📁 Project Structure

imageCaption/
├── backend/
│   ├── app.py              # Flask server + API routes + static file serving
│   ├── utils.py            # Model loading, VGG16 features, caption generation
│   ├── requirements.txt    # Python dependencies
│   └── model/
│       ├── best_model.h5   # Trained caption model weights
│       └── tokenizer.pkl   # Keras tokenizer (8485 vocab)
├── frontend/
│   ├── index.html          # React app (single-file, CDN-based)
│   └── style.css           # Dark UI with glassmorphism & animations
├── .gitignore
└── README.md

🔌 API

`POST /caption`

Upload an image and receive a generated caption.

curl -X POST -F "image=@photo.jpg" http://localhost:5000/caption

Response:

{ "caption": "a dog is running through the grass" }

`GET /health`

{ "status": "ok" }

👥 Team

Name	Roll No.	Role
Om Prakash Mishra	23051283	Tech Lead & Development Lead
Annada Shankar Maity	23051245	AI & ML
Kishalay Seren	2305626	AI & ML & Development
Suman Kumar Singha	2305822	R & D Lead
Priyanshu Dalei	2305626	R & D
Aryan Raj	2305607	R & D

📝 Notes

The model is trained on Flickr8k, a relatively small dataset. Caption accuracy improves significantly with larger datasets like Flickr30k or MS COCO.
The best_model.h5 file (~69 MB) is included in the repo. For very large models, consider using Git LFS.

Kishalay6/ImageCaptionGenerator