Text to image and Image to text

Generate image from text and vice versa

Project Description 

Developed a bidirectional AI system capable of generating images from text prompts and generating descriptive captions from images.

The project integrates two powerful pre-trained deep learning models:

stabilityai/stable-diffusion-2 for converting text prompts into high-quality AI-generated images.

nlpconnect/vit-gpt2-image-captioning for generating natural language descriptions from images.

The backend is built using Python Flask, which hosts the models and exposes REST APIs. The frontend is developed using HTML and JavaScript, enabling users to input text prompts or upload images and view results dynamically.

GPU acceleration is recommended for faster processing, although the system can operate on CPU with slower inference time.

Features 

• Generate realistic images from natural language prompts using Stable Diffusion.

• Generate meaningful captions from uploaded images.

• Smooth frontend-backend communication via HTTP requests.

• Optimized for GPU execution; supports CPU fallback.

• Simple and dynamic UI using HTML and JavaScript.

• Utilizes state-of-the-art diffusion and vision-language models.

Tech Stack

• Python 3.x

• Flask

• Transformers (Hugging Face library)

• PyTorch

• HTML5

• CSS3

• JavaScript

• HTML, JavaScript (Frontend)

System Requirements

• Python 3.8+

• Anaconda

• Libraries: numpy, pandas, scikit-learn, matplotlib

• Good to have GPU

Deliverables 

• Project Code

• Setup over video call

• WhatsApp support