This repository provides an implementation for GET3D with textual conditionning for converting text input into 3D models.
This project is divided project package in src textTo3DModelGen and multiple pipelines, each pipline stage is designed to handle a specific part of the process, from data ingestion to model training. Additionally, each stage includes a Jupyter notebook in the research folder that explains how thius pipeline works in detail.
To use this repository, clone it and install the required dependencies. Make sure you have Python installed, then run:
git clone https://github.com/mohamed-stifi/text-to-3D_Model_Generation.git
cd text-to-3D_Model_Generation
pip install -r requirements.txt- Download Blender
You can run different stages of the pipeline using command-line arguments. The available stages are:
python main.py --stage <stage_number>1: Run the Data Ingestion pipeline.2: Run the Data Rendering pipeline.3: Run the Text Embedding pipeline.4: Run the Data Split pipeline.5: Run the Training Model pipeline.6: Run all pipelines (not implemented yet).
The Data Ingestion stage collects and prepares raw data for further processing.
- Usage:
python main.py --stage 1 - Notebook:
research/01_data_ingestion.ipynb
The Data Rendering stage generates multi-view images of the 3D model.
- Usage:
python main.py --stage 2 - Notebook:
research/02_data_rendering.ipynb
The Text Embedding stage converts text data into embeddings suitable for model training.
- Usage:
python main.py --stage 3 - Notebook:
research/03_text_embedding.ipynb
The Data Split stage divides the dataset into training and testing sets.
- Usage:
python main.py --stage 4 - Notebook:
research/04_data_split.ipynb
The Training Model stage trains the model using the prepared data.
- Usage:
python main.py --stage 5 - Notebook:
research/05_training_model.ipynb
Each stage has a corresponding Jupyter notebook located in the research directory. These notebooks provide detailed explanations of the methodology, algorithms used, and implementation details for each stage.
01_data_ingestion.ipynb: Explains the data ingestion process. 02_data_rendering.ipynb: Details the data rendering step. 03_text_embedding.ipynb: Provides an overview of the text embedding technique. 04_data_split.ipynb: Describes how the dataset is split for training and testing. 05_training_model.ipynb: Walks through the model training process.
The core of the project is the textTo3DModelGen package. This contains the implementation of the different stages of the pipeline. and divided to multiple modules and sub-packages.
The entry point for running these pipelines is the main.py script, where you specify the stage to execute through the command-line arguments.