👕 AI Virtual Try-On System

An AI-powered Virtual Try-On (VTO) application that allows users to virtually wear clothing items using Computer Vision, Generative AI, and Large Language Models (LLMs).

The system combines image processing, segmentation, inpainting, and conversational AI to create an interactive virtual try-on experience.

---

🚀 Features

• 🧠 LLM-powered conversational agent
• 👕 Virtual clothing try-on
• ✂️ Human segmentation & garment extraction
• 🎨 AI inpainting integration
• 🔍 Garment image search support
• 💾 Memory-enabled conversations
• ⚡ LangChain + HuggingFace integration
• 🖼️ Real-time generated output images

---

🛠️ Technologies Used

• Python
• HuggingFace Transformers
• LangChain
• Computer Vision
• Image Segmentation
• Inpainting Models
• LoRA Fine-Tuning
• Generative AI

---

📦 Installation

1️⃣ Clone the Repository

git clone https://github.com/nhahub/NHA-289
cd NHA-289

2️⃣ Install Dependencies

pip install -r requirements.txt

---

⚙️ Environment Variables

Before running the project, configure the required environment variables:

import os

os.environ["VTO_LLM_MODEL"] = "Qwen/Qwen2.5-1.5B-Instruct"
os.environ["VTO_LLM_TOKEN"] = "your_huggingface_token"
os.environ["VTO_LORA_PATH"] = "ckpts/lora.safetensors"
os.environ["VTO_OUTPUT_DIR"] = "outputs"

---

▶️ Run the Application

python agent_runner.py

---

📁 Project Structure

NHA-289/
│
├── assets/                  # Input images
│   ├── clothes/             # Garment images
│   └── user_images/         # User uploaded images
│
├── ckpts/                   # LoRA fine-tuned weights
│
├── modules/                 # Core project modules
│   ├── agent_runner.py
│   ├── agent_vto.py
│   ├── image_search.py
│   ├── inpainting.py
│   ├── llm.py
│   ├── memory.py
│   ├── segmentation.py
│   └── vto_integration.py
│
├── outputs/                 # Generated try-on outputs
│
├── README.md
├── requirements.txt
└── final_project_test.ipynb

---

🧩 System Workflow

1. User uploads an image
2. Garment image is selected
3. Segmentation extracts clothing/body regions
4. AI inpainting generates the try-on result
5. LLM agent handles interaction and workflow orchestration
6. Final output is saved inside the outputs/ folder

---

🎯 Future Improvements

• Web application interface
• Multi-garment support
• Real-time try-on
• Better diffusion-based generation
• Mobile deployment
• Personalized fashion recommendations

---

🤝 Contributing

Contributions, suggestions, and improvements are welcome.

Feel free to fork the repository and submit a pull request.

---

📜 License

This project is intended for educational and research purposes.

🎥 Demo