ReModel

This was a project we designed for the Georgia Tech Hacklytics 2025 Hackathon.

Link to our DevPost.

Models

Current

Our current stack includes the following models:

Model	Origin	Link	Purpose and Details
GPT-4o	OpenAI	OpenAI
GPT-4o mini	OpenAI	OpenAI
StableDesign		Github & Replicate API	Comprised of multiple other layered models including
YOLO 8m	Ultralytics	GitHub & Docs
Segment Anything Model (SAM) VIT-H	Meta Research	GitHub

Past and Planned

During previous stages of our development and/or in the intended designs we had but weren't able to implement yet, we include the following models in addition to the current:

Model	Origin	Link	Purpose and Details
Contrastive Language-Image Pre-Training (CLIP) Commit dcba3cb	OpenAI	GitHub
OmniGen	VectorSpaceLab	GitHub & Replicate & Hugging Face
StableDiffusion

Running Locally

We recommend setting up a conda or ve`nv environment to run this.

1. Python Requirements

Within our pip requirements includes various packages for api calls and model downloads (such as for openai, CLIP, and replicate).

pip install -r requirements.txt

2. Selenium Setup

This is used for our content matching phase for product matching purposes.

Download Page: https://googlechromelabs.github.io/chrome-for-testing/#stable

3. Local Model Setup

Ultralytics's You Only Look Once 8m (Yolo8m) and Meta (Facebook) Research's Segment Anything Model (SAM) models are required for our content extraction phase. They're used for furniture and decor item identification within our generated design plans.

Model Name	Direct Download Link	Docs and Details
YOLO8m	Ultralytics Direct Download	Ultralytics YOLO8m Docs
VIT-H SAM	Meta's Direct Download	Facebook Research's Github SAM Docs

4. Running the Program

streamlit run app.py