Mirror Mirror AI
A smart mirror that can help you decide if you're comfortable with your outfit.
Abstract
Mirror Mirror AI is an innovative smart mirror that redefines the daily dressing experience.
Using a custom-trained YOLO model on a dataset we created and labeled ourselves, the mirror identifies clothing items in real time. It then recognizes the user standing in front of it with DeepFace-based facial recognition, ensuring that recommendations are personal and unique for each individual.
At the core of the system lies our adaptive recommendation engine, a feedback-driven model that we independently designed and developed. By learning from user responses (“too hot,” “too cold,” “comfortable”), the system evolves over time to deliver increasingly accurate and personalized outfit suggestions.
Beyond weather-matching outfits, Mirror Mirror AI creates a seamless fusion between fashion and technology, turning a simple mirror into an intelligent companion that answers the timeless question: “Who’s the most stylish of them all?”
Curious?
Meet the smart mirror everyone will talk about
Live Examples
Real-Time User Recognition
Watch as the mirror instantly identifies different users and switches between their personal profiles, remembering individual preferences and style history for each person.
Smart Clothing Detection
Our custom YOLO model accurately identifies clothing items in real-time, analyzing colors, patterns, and styles to understand what you're wearing or considering.
Weather-Adaptive Suggestions
See live recommendations that adapt to current weather conditions, seamlessly blending comfort with style for the perfect daily outfit.
Natural Voice Interaction
Experience hands-free communication with speech-to-text input and text-to-speech responses, making the mirror accessible and intuitive for everyone.
Learning from Feedback
Watch the AI evolve in real-time as it learns from user feedback like "too hot" or "perfect," continuously improving its recommendations for each individual.
Research Results & Analysis
Al Model Feature Importance
User history and temperature are the most influential factors for predictions.
System Response Time Breakdown
Real-time performance: Complete analysis delivered in under 2 seconds.
Model Accuracy Evolution
Demonstrating how the Al system learns and improves from user feedback, evolving from basic rules to personalized intelligence.
User Preference Evolution: Hot -> Cold Sensitivity
Real-time adaptation: Watch how the Al detects and responds to changing user preferences over time.
Key Contributions
Custom Dataset & Model Training
We created and labeled our own dataset and used it to train a YOLO model tailored to clothing recognition. This significantly improved accuracy and ensured the system was built on data directly relevant to our use case, rather than relying on generic pre-trained sets.
Adaptive Recommendation Engine
At the heart of the project lies a feedback-driven recommendation system we designed and developed ourselves. By learning from user inputs (“too hot,” “too cold”), the engine continuously refines outfit suggestions, making the experience more personal and accurate over time.
Personalized User Interaction
The mirror recognizes the individual standing in front of it through DeepFace-based facial recognition. Each user receives unique, profile-specific recommendations, ensuring that personalization is not just a feature but a defining characteristic of the system.
Accessible & Natural Feedback
With speech-to-text for input and text-to-speech for output, user interaction becomes seamless and inclusive. This lowers barriers for people with visual impairments and makes the system easy and intuitive for everyday use.
Bridging Fashion and Technology
Beyond the technical aspects, Mirror Mirror AI demonstrates how artificial intelligence can be meaningfully integrated into daily life, turning a simple mirror into an evolving AI companion that blends practicality, creativity, and style.
Future Work
- Enhanced recognition accuracy – Further improving the YOLO model and dataset to better distinguish between visually similar garments and users.
- Style-based recommendations – Moving beyond weather and comfort by incorporating each user’s fashion taste and trends.
- User-friendly interface – Developing a polished UI with dashboards, outfit history, and style insights.
- Scalability & integration – Adapting the system for wider platforms such as mobile apps or smart home devices.
Technical Details
Methodology
Our project explores how computer vision and machine learning can enhance the everyday act of getting dressed.
We developed Mirror Mirror AI, a smart mirror that identifies clothing in real time, recognizes the individual user standing in front of it, and delivers personalized outfit recommendations.
The system is powered by a custom-trained YOLO model for clothing detection, DeepFace for user recognition, and an adaptive feedback-driven recommendation engine we designed ourselves.
Experimental Setup
We created and labeled a custom dataset to train the YOLO model, ensuring accurate detection across multiple clothing categories.
The mirror registers users with facial recognition and maintains personalized profiles. Recommendations are then generated by combining live weather data (via OpenWeatherAPI) with user-specific history and feedback (“too hot,” “too cold”).
A speech-to-text interface allows users to provide feedback naturally, while a text-to-speech engine delivers verbal recommendations.
Results Analysis
Our findings show that the system successfully identifies garments in real time, recognizes users, and provides increasingly accurate outfit suggestions with each interaction.
The adaptive recommendation engine improves personalization as feedback accumulates, making the mirror not only a practical assistant but also an evolving AI companion.
The results demonstrate how machine learning can bridge the gap between fashion, accessibility, and technology, offering value to everyday users as well as people with visual impairments.
Installation Guide
Step 1: Clone the repository
git clone https://github.com/iLihiS/git-dinamic-page.git
cd git-dinamic-page
Step 2: Install dependencies
pip install -r requirements.txt
Step 3: Run the main script
python main.py