👗 AI Fashion Stylist

An AI-powered fashion recommendation system that suggests complementary clothing items and styling ideas based on an uploaded image.
Built using DeepFashion2 (Kaggle version), CLIP embeddings, and ResNet50 classification, it provides occasion-aware, color-aware, and category-smart outfit recommendations.

---

🌟 Features

✅ Automatic Clothing Detection

• Detects whether an uploaded image is Topwear, Bottomwear, Dress, or Outerwear using a trained ResNet50 model.

✅ Smart Outfit Suggestions

• Suggests complementary clothing items (e.g., Top → Bottomwear, Bottom → Topwear).
• Uses CLIP embeddings and FAISS similarity search for top 5 visually similar items.

✅ Color Analysis

• Extracts dominant colors using K-Means clustering and provides human-readable color names.
• Generates color harmony tips and contrast-based fashion advice.

✅ Occasion-Aware Styling

• Provides dynamic styling tips based on detected color tone and outfit type — for Casual, Formal, Party, Streetwear, or Festive occasions.

✅ Local Execution

• Works fully offline once the dataset and models are set up on your local machine.

---

🧠 Tech Stack

• Python 3.10
• PyTorch / torchvision
• OpenAI CLIP
• FAISS (Facebook AI Similarity Search)
• Streamlit
• scikit-learn / OpenCV / Pillow
• Kaggle: DeepFashion2 Original Dataset

---

🗂️ Project Structure

AI-Fashion-Stylist/
│
├── DeepFashion2/
│   ├── deepfashion2_original_images/
│   └── img_info_dataframes/        # train.csv, validation.csv, test.csv
│
├── data_subset/                    # Prepared smaller dataset for training
│
├── models/
│   ├── cloth_classifier.pth        # Trained ResNet50 model
│   ├── fashion_index.faiss         # FAISS index built from CLIP embeddings
│   └── image_paths.pkl             # List of images used in FAISS index
│
├── scripts/
│   ├── prepare_dataset.py          # Dataset pre-processing
│   ├── train_classifier.py         # Model training
│   ├── build_faiss.py              # CLIP + FAISS builder
│   └── color_utils.py              # Color extraction utility
│
├── demo_app.py                     # Streamlit application
└── README.md                       # You're reading it!

---

⚙️ Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/ritup04/AI-Fashion-Stylist.git
cd AI-Fashion-Stylist

2️⃣ Install Dependencies

pip install -r requirements.txt

⚠️ Make sure you’re using Python 3.10 for best compatibility with PyTorch and FAISS.

---

🧩 Dataset

This project uses the DeepFashion2 dataset available on Kaggle.

📦 Dataset Link:
DeepFashion2 Original with Dataframes (Kaggle)

After downloading, extract it into:

AI-Fashion-Stylist/DeepFashion2/

⚠️ The dataset is not included in this repository due to its large size.

---

🧩 Run the Complete Pipeline

🧱 Step 1 — Prepare Dataset

python scripts/prepare_dataset.py

🧠 Step 2 — Train Classifier

python scripts/train_classifier.py

🧮 Step 3 — Build FAISS Index

python scripts/build_faiss.py

💅 Step 4 — Launch the AI Stylist App

streamlit run demo_app.py

---

🖼️ Demo Workflow

Step	Description
🖼️ Upload	Upload an outfit image (topwear, bottomwear, dress, etc.)
🧠 Detection	The model classifies the outfit type
🎨 Color Extraction	Extracts dominant colors & harmony
🛍️ Similar Items	Displays top 5 similar or complementary outfits
💡 Styling Tips	Gives dynamic, occasion-based fashion advice

---

🧾 Authors

👩‍💻 Ritu Pal
🎓 B.Tech CSE (AI-ML), Adani University
📧 ritupal1626@gmail.com
💼 GitHub: ritup04

👩‍💻 Helly Khambhatwala
🎓 B.Tech CSE (AI-ML), Adani University
📧 helly9328@gmail.com
💼 GitHub: helly1408