---
tags:
- text-to-image
- creative
- control-image
- stable-diffusion
pinned: false
datasets:
- ClaireOzzz/All_Images
- ClaireOzzz/Porcelain
license: mit
language:
- en
---

# SUTDxSUNS Spatial Project: A 2.5D Control Generator 

![image/png](https://github.com/ClaireOzzz/ImageURLhost/blob/main/Prototype%20Tests.png?raw=true)

## Status

Currently ongoing - will be completed by the 17th of December

## Demo

Try out the prototype on [COLAB](https://colab.research.google.com/drive/18FBtOWcf9jgLGV-p-ATPNtCHus2yWe-W?usp=sharing)

## Project Description

This project was for a short-term design studio that required applied deep learning and linking of existing models for spatial design. Special thanks to SUNS for providing us with data.
The controlnet training model and generation code is credited to fffiloni, while the depthmap generation to GLTF model is credited to Radamés Ajna.

[Training model](https://huggingface.co/spaces/fffiloni/train-dreambooth-lora-sdxl)
[Generation model](https://huggingface.co/spaces/fffiloni/sdxl-control-loras)
[3D Depth model](https://huggingface.co/spaces/radames/dpt-depth-estimation-3d-obj)

## Pipeline

![image/png](https://github.com/ClaireOzzz/ImageURLhost/blob/main/New%20Pipeline.png?raw=true)


### Key Features

- **User-Friendly Training Interface for non-developers**
  
- **Text-to-Image Generation with control** 

- **Specific Training Data** 

- **2.5D Visualisation** B