🚘 Lane Detection using U-Net

This project implements road lane detection using a U-Net deep learning architecture.
It provides training code, prediction outputs, and an interactive Gradio UI for real-time testing on images, videos, and webcam feeds.

---

📂 Project Structure

lane_detection_using_unet/
│── app.py                     # Gradio app for demo
│── final_project.ipynb         # Training & evaluation notebook
│── requirements.txt            # Dependencies
│── README.md                   # Project documentation
│── new_dataset/                # Training & prediction samples
│   ├── training/
│   ├── prediction/
│   └── prediction_refined/
│── old_dataset/                # Tu Simple lane prediction dataset
│   ├── TUSimple/
├         ├── train_set
├         ├── test_set
├         ├── test_label
│── .gitignore                  # Keeps repo clean

---

🔧 Installation

1️⃣ Clone the repository

git clone https://github.com/your-username/lane_detection_using_unet.git
cd lane_detection_using_unet

2️⃣ Install dependencies

pip install -r requirements.txt

---

🚀 Run the App

python app.py

This launches a Gradio web app with three modes:

• 🖼️ Image Detection → Upload an image
• 🎬 Video Detection → Upload a video
• 📹 Live Webcam → Real-time lane detection

---

🖥️ User Interface

Here’s the Gradio-based UI of the project:

---

🖼️ Prediction

Sample output of lane detection (green overlay = detected lanes):

---

📊 Results

• Model: U-Net with skip connections
• Input size: 128x256
• Framework: TensorFlow / Keras
• Accuracy: Achieved stable lane segmentation on road datasets

---

🎥 Demo Video

👉 Watch Demo Video

---

📌 Notes

• The model is trained on a TUSimple lane detection dataset.
• You can retrain the model using the final_project.ipynb notebook.

---

📜 License

This project is released under the MIT License.