Representing shapes as Latent Codes

This project was conducted as part of the Advanced Deep Learning for robotics course at the Technical University of Munich.

📌 Overview

This project focuses on deep learning techniques for robotics applications, including:

• Latent space optimization
• 3D object representation
  The implementation extends the DeepSDF framework and explores improvements in multi-class training.

📂 Repository Structure

project_root
┣ configs/
┃ ┣ hyperparameters.yaml
┃ ┣ settings.yaml
┣ multi_class/
┃ ┣ data/
┃ ┣ trained_models/
┃ ┣ visualization/
┣ scripts/
┃ ┣ data_manipulation/
┃ ┣ evaluation/
┃ ┣ helpers/
┃ ┣ models/
┃ ┣ training/
┃ ┣ utils/
┣ README.md
┣ requirements.txt

🚀 Setup Instructions

Follow these steps to set up and run the project:

1️⃣ Create a Virtual Environment

Run the following command to create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On macOS/Linux
venv\Scripts\activate     # On Windows

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Train and Evaluate the Model

python main.py --mode multi_class  # Multi-class training
python main.py --mode single_class # Single-class training