Text_to_image_generation_with_Dreambooth

This project transforms textual descriptions into high-quality images & involves using machine learning models and neural networks to generate professional attire images from regular images of women. It employs the runwayml/stable-diffusion-v1-5 model and fine-tunes it using Dreambooth on the established dataset of 1000 images. The project leverages numerous tools and frameworks including Transformers, PyTorch, and Hugging Face’s diffusers.

Table of Contents

• Prerequisites
• Installation
• Usage
• Model Training
• Data Preparation
• Results
• Contributing
• Contact

Prerequisites

Before you begin, ensure you have met the following requirements:

• Python 3.10 or higher installed on your machine
• Access to Google Colab (recommended for GPU support)
• A connected and configured Google Drive to access datasets

Installation

1. Clone the repository:

git clone https://github.com/anou1234/Text_to_image_generation_with_Dreambooth.git

2.Navigate to the project directory:

cd myproject

3. Install the required dependencies:

pip install transformers torch bertviz datasets
git clone https://github.com/ShivamShrirao/diffusers
cd diffusers/
pip install -e .
pip install bitsandbytes
cd ../
pip install -r requirements.txt

Usage

1. Upload your image dataset. Uploading by mounting Google Drive is recommended. Adjust paths as appropriate:

import os
import shutil
from google.colab import drive
drive.mount('/content/drive')
source_dir = "/content/drive/MyDrive/woman" #dataset folder
destination_dir = "/content/images"
num_files_to_copy = 100

Fine-tune the model using provided scripts in the examples/dreambooth directory.

Model Training

Training a customized model involves the following steps:

• Setting up an accelerated environment for efficient computation.
• Using the notebook gen_ai_dreambooth.ipynb and following instructions to run the cell blocks for model training and fine-tuning.

Data Preparation

Prepare your data as follows:

• Ensure images are stored correctly in your dataset folder. Eg. For a 'woman' token dataset, the images should be named as "woman_1" , "woman_2" , and so on.
• Modify file paths as required in scripts to point to directories storing your data.

Results

Once the model is trained, generate and visualize your results. Evaluate the model's effectiveness in transforming images and enhancing attributes specified during training.

Contributing

Contributions are always welcome! Please fork the repository and create a pull request. For major changes, please open an issue first to discuss what you would like to change.

Contact

• Author: Anoushka Srivastava
• Email: anoushkathegreat28@gmail.com/anoushka.srivastava.21cse@bmu.edu.in
• GitHub: anou1234