Fashion.AI- Personalized Outfit Recommendation

Abstract

In the last few years, AI has slowly started changing how the fashion industry works, especially when it comes to giving people outfit ideas that feel more personal. For this project, we made a fashion assistant that helps suggest clothes by breaking the job into smaller parts. It uses things like image search, color picking, and some smart tools to figure things out. When someone uploads a picture, the system checks what type of clothing is in it using a deep learning method called Faster R-CNN (basically, it helps the system \"see\" the clothes). Then, it tries to figure out the main color of the outfit using K-means clustering, which just means it groups similar colors together. After that, it sends this info, along with whatever style the person typed in, to ChatGPT. That’s where the outfit suggestions come from—it uses all that input to recommend clothes that fit the style. To help users picture the results, the system also pulls similar images from the Pexels site. The whole thing works through a simple and friendly interface that walks you through each step.

Introduction

I. Overview
Fashion is a key form of self-expression, and many people seek help in making style decisions. Traditionally, advice came from stylists or fashion media. Now, with AI and computer vision, it's possible to create virtual fashion consultants that provide personalized outfit suggestions based on images and style prompts. This project presents such a system: a smart fashion assistant that analyzes uploaded clothing images and style prompts to generate complete outfit ideas, including visual references.

II. Problem Statement

Current fashion advice platforms lack:

• Personalization to individual wardrobes and styles,

• Visual suggestions that match real clothes,

• Real-time interaction or feedback.

There's a need for an AI-based system that:

• Detects clothing from user-uploaded photos,

• Analyzes color,

• Understands user prompts (e.g., “party outfit”),

• Provides simple, customized outfit suggestions with matching visuals.

III. Goals

• Main Goal: Create an AI fashion assistant that analyzes clothes and suggests matching outfits with visual previews.

• Secondary Goals:

  • Clothing detection via object recognition,

  • Color analysis to identify dominant hues,

  • ChatGPT integration for text-based recommendations,

  • Image fetching from Pexels for visual suggestions,

  • User-friendly interface to make the system accessible and simple.

IV. Literature Review

• Technologies like Faster R-CNN (Ren et al.) and DeepFashion datasets (Liu et al.) aid in clothing detection.

• K-means clustering is effective for color extraction.

• Prior systems often lack real-time, personalized suggestions or integration with NLP tools like ChatGPT.

• Some models, like Fashion++ and Style2Vec, handle style but don't offer dynamic user interaction or complete outfit generation.

• This project fills the gaps by combining object detection, color analysis, prompt-driven AI, and image retrieval in a real-time, interactive fashion system.

V. Unique Contributions

This system:

• Extends object detection with real-time outfit recommendations,

• Uses ChatGPT for interpreting natural language and generating style ideas,

• Adds visual feedback using Pexels API,

• Works with minimal user input (an image + a prompt),

• Provides personalized and interactive styling, unlike prior static systems.

VI. Methodology

1. GUI Module: Users upload an image and input a text prompt (e.g., “summer picnic outfit”).

2. Detection Module: Uses Faster R-CNN to identify clothing items with >50% confidence.

3. Color Analysis Module: Applies K-means clustering to detect the dominant clothing color.

4. ChatGPT Module: Combines item + color + user prompt to generate outfit suggestions in text.

5. Image Fetching Module: Queries Pexels for images matching suggested items.

6. Result Display: Shows the uploaded image with detected clothing and color labels, plus outfit suggestions and photos.

VII. Results & Discussion

• Detection Accuracy: 85% for typical clothes (lower for complex or unique pieces).

• Color Detection Accuracy: >90% for solid colors; less accurate for patterned items.

• Outfit Suggestion Relevance: Highly accurate when users give clear prompts.

• Visual Output: High-quality, helpful images from Pexels enhance understanding.

Key Findings

• Clear prompts improve results.

• Dominant color detection is crucial.

• Visual aids increase user satisfaction.

Limitations

• Struggles with uncommon or mixed-pattern clothes.

• Visual search may fail for rare items.

• Complex styles may be misclassified.

VIII. Future Scope

1. Expanded Dataset: Broader style and cultural clothing samples for more diverse suggestions.

2. Improved Personalization: Learning user preferences over time for better recommendations.

3. Cloud Deployment: For better scalability and faster performance under high usage.

Conclusion

This fashion system mixes a few smart tools to help people get outfit ideas. It looks at what’s in the photo, figures out the main clothing and its color, and then gives outfit suggestions based on what the user wants. It even shows pictures from the web to make it more visual.It works pretty well most of the time, especially if the user gives a clear request. Some clothes with patterns or weird styles confuse it a bit. Still, it’s a helpful start, and with more updates, it could become even better at giving personal style advice

References

[1] S. Ren, K. He, R. Girshick, and J. Sun, \"Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,\" IEEE TPAMI, vol. 39, no. 6, pp. 1137–1149, 2017. [2] Z. Liu, P. Luo, S. Qiu, X. Wang, and X. Tang, \"DeepFashion: Powering Robust Clothes Recognition and Retrieval,\" in CVPR, 2016. [3] F. Pedregosa et al., \"Scikit-learn: Machine Learning in Python,\" J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011. [4] OpenAI, \"GPT-4 Technical Report,\" OpenAI, 2023. [Online]. Available: https://platform.openai.com/docs [5] S. Jagadeesan and R. V. Babu, \"Fashion++: Minimal Editing of Fashion Images to Improve Outfit Compatibility,\" in ICCV, 2019. [6] J. Song et al., \"Deep Style Transfer for Coherent Visual Fashion Recommendation,\" IEEE Trans. Image Process., vol. 29, pp. 2287–2300, 2020. [7] J. McAuley, C. Targett, Q. Shi, and A. van den Hengel, \"Image-based Recommendations on Styles and Substitutes,\" in SIGIR, 2015. [8] Y. Hu, X. Yi, and L. S. Davis, \"Collaborative Fashion Recommendation: A Functional Tensor Factorization Approach,\" in ACM MM, 2015. [9] K. Yamaguchi et al., \"Parsing Clothing in Fashion Photographs,\" in CVPR, 2012. [10] H. Han, S. Wu, and J. Yan, \"Style2Vec: Representation Learning for Fashion Items from Style Sets,\" in WWW, 2017. [11] L. Bossard et al., \"Apparel Classification with Style,\" in Asian Conference on Computer Vision, 2013. [12] H. Chen, A. Gallagher, and B. Girod, \"Describing Clothing by Semantic Attributes,\" in ECCV, 2012. [13] Pexels, “Free Stock Photos & API,” [Online]. Available: https://www.pexels.com/api/ [14] S. Yarosh and K. Ryokai, \"Supporting Parent–Child Interaction in Joint Media Engagement,\" IUI, 2011. [15] A. Yadav, M. Chauhan, and R. Meena, \"A Comprehensive Review on Fashion Recommender System Using Machine Learning,\" SSRN, 2023.

Copyright

Copyright © 2025 Krushabh Roy, Hasnain Patil, Wahid Shaikh, Sara Kurekar, Prof. Reena Pagare. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.