Portfolio | Renosmarter.ai

Visit Web

Overview

RenoSmarter.ai is an innovative platform designed to transform the home renovation process. By integrating AI with user-friendly design tools, RenoSmarter.ai empowers homeowners, renovation enthusiasts to make informed decisions, design dream spaces, and manage renovation costs effectively.

Lack of Accessible Information

Homeowners often lack basic renovation knowledge, leading to costly mistakes and unsatisfactory results.

Challenges in Envisioning

Difficulty in visualizing renovation outcomes results in inefficient designs that don't meet expectations.

Inconsistent Cost Estimations

Inaccurate material cost estimates cause budget overruns and make it hard for homeowners to adjust projects within financial limits.

Customer Challenges

Customer Research

During the initial weeks, our team was tasked with interviewing as many people as possible, including those who are house owners or apartment owners. We came up with two personas after we finished all the interviews with those potential customers.

What did we ask?

1.Customer’s background: If interviewee is a first time home buyer?

2.Discovery of Pain Points: How long did interviewee live in their home until they realized the workflow pain points?

3.Current Home Satisfaction: What do you like or do not like about your current home?

4.Barriers to Renovation: What is preventing you from addressing those home issues?

5.Vision for ideal Renovation Project: What does an ideal end to end renovation project look like for you?

6.Additional Challenge: Are there any other challenges that you would like to share regarding home renovation?

Personas

AI Model

Purpose: To generate personalized responses and provide design guidance based on user input and preferences. The model interprets text queries and can also respond with both text and relevant images, enhancing user interaction.
Model Characteristics: GPT-4o is a highly advanced LLM known for its ability to generate human-like text and understand nuanced context. It can handle complex conversations, providing tailored design recommendations and offering visual aid suggestions (such as related images from a database) based on the user's preferences (e.g., traditional, modern, farmhouse). GPT-4o can also integrate relevant image outputs, working in conjunction with Stable Diffusion, to enhance design experiences.

GPT-4o (Text-to-Text, Text-and-Image)

Stable Diffusion (Image-to-Image)

Purpose: To create realistic design visualizations by transforming or generating images based on user-provided input or style preferences. The model can refine design ideas or offer alternative visual representations of concepts like room layouts, materials, or stylistic elements.
Model Characteristics: Stable Diffusion is a cutting-edge generative model for creating highly detailed images from textual or image inputs. Its strength setting can be adjusted from 0 to 1, with 0 offering minimal changes and adhering closely to the input image, while 1 allows for more creative, out-of-the-box interpretations. This model also supports "negative prompts," which help the model avoid unwanted design elements or styles during the generation process.

AI Input and AI Output

AI Input

AI Output

Material Data: Includes detailed information about building materials (e.g., concrete, wood, tile) such as size, color, and price per unit. Visual data (e.g., images) for each material is also fed into the model.
User Preferences: User inputs such as style preferences (e.g., traditional, farmhouse, modern), budget constraints, and specific material choices are collected.
Pricing Data: Collected data on material costs helps generate accurate estimates for renovation projects.
Design Guidelines: Predefined home renovation guidelines, including kitchen layout standards, are integrated to enhance suggestions during the design process.
Image Inputs: User-uploaded images (e.g., current home design or layout) are processed using Stable Diffusion for image enhancement, transformation, or design visualization. Images are converted to string/binary format for efficient processing and transferred via API for Stable Diffusion's image-to-image operations.

Personalized Design Recommendations: GPT-4o generates custom design guidance based on user preferences (e.g., traditional, modern) and material choices. It provides suggestions for color palettes, furniture arrangement, and other stylistic elements that align with the user's desired aesthetic.
Material Visualization: Stable Diffusion generates visual representations of selected materials (e.g., flooring, countertops) and their application within a space. The user can preview how different materials and styles will look in their home before making final decisions.
Cost Breakdown: The system provides a transparent breakdown of estimated costs for materials and labor, using data inputs such as regional labor rates and material pricing to help users understand their budget and explore cost-effective alternatives.
Visual Design Previews: Using the image-to-image capabilities of Stable Diffusion, users can upload photos of their current home or layout, and the model will generate modified versions that incorporate new design elements, materials, or styles based on user preferences. The strength parameter (0 to 1) allows users to control how closely the final image aligns with their original input or how creatively the model reinterprets their design vision.

Data Pipeline

Step 1: Data Collection

Material Database:

Data is collected from an established database of building materials sourced from providers like Floor and Decor. This includes materials for flooring, cabinets, countertops, and backsplash with detailed attributes such as material type (e.g., concrete, wood, tile), size, price per unit, color, and description.
Visual data such as images and descriptions of each material is also collected to provide users with comprehensive information.

Labor Rate Data:

General labor rate ranges for different renovation tasks are collected to help provide accurate pricing and budget estimates.

Renovation Best Practices:

Home renovation guidelines and best practices from relevant documents (e.g., PDFs) are incorporated into the knowledge base. This data helps guide users through their renovation process with expert advice and tips.

Step 2: Data Preprocessing

Material Data Normalization:

Collected material data (e.g., size, price per unit, color) is cleaned, organized, and standardized to ensure uniform data formats.
This data is categorized by item type (flooring, cabinets, countertops, backsplash) and material (concrete, wood, tile).

Pricing Data Structuring:

Labor rate data is segmented by task type (e.g., flooring installation, cabinetry) and normalized based on region-specific variations, allowing for accurate pricing suggestions during the renovation process.

Knowledge Integration:

Renovation guidelines and best practices are categorized into different sections (e.g., materials, layout advice) and prepared for use in the chatbot's suggestions and recommendations.

Step 3: AI Model Preprocessing

Chatbot (GPT-4o)

User preferences such as style, budget, and material choices are processed by GPT-4o to deliver personalized, conversational guidance.

The model incorporates renovation best practices and layout data (e.g., “kitchen in the corner”) to generate contextual design suggestions and practical advice.

The chatbot uses gathered data to provide tailored responses and engage users with empathetic, actionable suggestions.

Image generation(GPT-4o/Stable Diffusion)

The design model leverages collected material data (e.g., images, descriptions) and user preferences to generate visually appealing design suggestions.

Different room elements (e.g., flooring, cabinets) are integrated into visualizations based on user input, offering realistic previews of the renovation.

Step 4: Output Generation and Feedback Loop

Personalized Design Suggestions:

The chatbot generates real-time design suggestions tailored to the user’s preferred style (Traditional, Transitional, Farmhouse, Modern) and material preferences.

Transparent Pricing and Material Recommendations:

The chatbot offers detailed breakdowns of material prices, labor costs, and budget-friendly alternatives. Users receive upfront estimates based on their material choices and regional labor rates.

User Feedback Integration: