You are on page 1of 4

AI Interior Designer

Project Proposal

Supervisor
Junaid Ahmed

Submitted by

Rehman Shabbir
BSSE-E2-20-59

Department of Computer Science,


University of Sahiwal.

[07-03-2024]
1. Project Title
AI Interior Designer

2. Introduction
The project aims to harness the power of artificial intelligence and image
processing to create a mobile application that can transform the interior of a
room based on user inputs. While there are applications that allow users to
visualize interior design changes, few, if any, leverage the power of AI and
image processing to generate images based on textual prompts. This project
seeks to fill this gap by developing a unique application that combines the
power of AI, image processing, and user creativity.

3. Objective
This project aims to develop a mobile application that uses artificial
intelligence and image processing to transform the interior of a room based on
user inputs. The application will leverage the Stable Diffusion’s Img2Img
pipeline and Transformers to generate visually stunning images from textual
prompts.

4. Problem Description
The challenge is to develop a mobile application that can accurately interpret
textual prompts and generate high-quality images that reflect the desired
changes in interior design. This involves understanding the nuances of
language, the principles of design, and the technical aspects of image
generation.

5. Methodology
The project will utilize the StableDiffusionImg2ImgPipeline, which is pre-
trained on a vast dataset of high-quality images. This pipeline leverages the
capabilities of Transformers and Stable Diffusion to turn textual prompts into
remarkable visual content. The pipeline will be integrated into a mobile
application, providing users with an intuitive interface to input their design
ideas.

6. Project Scope
This project has applications in the field of interior design, architecture, and
home improvement. It can be used by professional interior designers for
conceptualization and visualization, homeowners for DIY projects, or even
real estate agents for virtual staging.

7. Feasibility Study
Risks Involved: The main risk is the accurate interpretation of textual
prompts into visual content. This will be mitigated by using a robust AI model
trained on a vast dataset.
Resource Requirement: The project will require computing resources for the
development and testing of the mobile application, as well as access to the
StableDiffusionImg2ImgPipeline.

8. Solution Application Areas


The application could also be beneficial for furniture retailers and real estate
developers. Furniture retailers can use it to showcase how their products can
fit into various interior design styles. Real estate developers can use it to
provide potential buyers with a visualization of how the interior of the
properties can be customized.

9. Tools/Technology
 The project is a web application that uses artificial intelligence and image
processing to transform room interiors based on user inputs.
 Tools and technologies required include:
 Web application development platform
 AI modeling tools
 StableDiffusionImg2ImgPipeline
 Database management system
 Web hosting services
 Version control system
 Testing tools

10. Expertise of the Team Members


The team members have a strong background in AI, image processing, and
mobile application development, making them well-equipped for the
successful completion of this project

10. Milestones
 Requirements Gathering: Define the specific requirements of the
web application, including user interface, functionality, and
performance needs.
 Design Phase: Create detailed design documents outlining how the
application will work from a technical standpoint.
 Development of Web Application Interface: Build the user interface
of the web application, ensuring it is user-friendly and intuitive.
 Integration of the StableDiffusionImg2ImgPipeline: Incorporate the
StableDiffusionImg2ImgPipeline into the web application, enabling
the generation of images from textual prompts.
 Initial Testing: Conduct initial testing to identify any bugs or issues
that need to be addressed.
 User Feedback and Refinement: Release the application to a select
group of users for feedback, and make necessary refinements based on
their input.
 Final Testing: Perform final testing to ensure the application is
working as expected.
 Project Completion: Launch the application to the public, and provide
ongoing support and updates as needed.

11. References
1. Chen, T. Q., et al. (2021). Generative Models and the Stabilizing Diffusion.
ICLR 2021.
2. Grathwohl, D., et al. (2021). Diffusion Models Beat GANs on Image
Synthesis. ICLR 2021.
3. A.I., S. (n.d.). Stable diffusion public release. Retrieved from
https://stability.ai/blog stable-diffusion-public-release
4. Alaluf, Y., Tov, O., Mokady, R., Gal, R., & Bermano, A.H. (2022).
Hyperstyle: Stylegan inversion with hypernetworks for real image editing.
arXiv:2111.15666 [cs].
5. Brock, A., Donahue, J., & Simonyan, K. (2019). Large scale GAN training for
high fidelity natural image synthesis. ICLR 2019.
6. Chen, T. Q., Li, X., Grosse, R. B., & Duvenaud, D. (2018). Isolating sources
of disentanglement in variational autoencoders. NeurIPS, 5594-5603.
7. Karras, T., Aila, T., Laine, S., & Lehtinen, J. (2018). Progressive growing of
GANs for improved quality, stability, and variation. arXiv preprint
arXiv:1710.10196.
8. Kingma, D. P., & Welling, M. (2013). Auto-encoding variational bayes. arXiv
preprint arXiv:1312.6114.
9. Radford, A., Metz, L., & Chintala, S. (2016). Unsupervised representation
learning with deep convolutional generative adversarial networks. ICLR 2016.
10. Wu, J., Zhang, C., Xue, T., Freeman, B., & Tenenbaum, J. (2019). Learning a
probabilistic latent space of object shapes via 3D generative-adversarial
modeling. NeurIPS, 82-92.
11. Hjelm, R. D., Fedorov, A., Lavoie-Marchildon, S., Grewal, K., Trischler, A.,
& Bengio, Y. (2019). Learning deep representations by mutual information
estimation and maximization. ICLR 2019.
12. McDermott, S. D., & Mahoney, M. W. (2021). Adaptive importance sampling
for diffusion-based generative models. arXiv preprint arXiv:2102.02760.
13. Brock, A., Donahue, J., & Simonyan, K. (2018). Large scale GAN training for
high fidelity natural image synthesis. ICLR 2018.

You might also like