TuSimple-Lane-Detection-with-ResNet-UNet-Focal-Loss

📌 Features:

✔️ UNet segmentation model with a ResNet-50 Encoder

✔️ Optimized loss function using Focal Loss + Dice Loss

✔️ Data augmentation techniques for better training

✔️ Evaluation metrics such as IoU, Dice Score, and Precision-Recall Curve

✔️ Training & testing pipeline, including model saving and loading

🚀 How to Run:

1️⃣ Download the dataset (Kaggle TuSimple Dataset)

2️⃣ Train the model by running train.py

3️⃣ Test the model using test.py to view evaluation metrics

📊 The test results are saved as a Precision-Recall (PR) curve and prediction images.