Rui tests

[Rui-Pedro-Pires]

This pull request introduces significant changes to the lane detection project by adding new functionalities and improving existing ones. The most important changes include the addition of a dataset class for loading and processing data, a deep learning model for lane segmentation, a main script for inference and model export, a training script, and utility functions for visualization.

Dataset and Data Processing:

• Added TuSimpleDataset class in src/Dataset.py to handle loading and processing of the TuSimple dataset, including image transformations and binary label generation.

Model Definition:

• Introduced LaneSegmentationModel class in src/Model.py using DeepLabV3 with a ResNet-50 backbone for lane segmentation tasks.

Inference and Model Export:

• Implemented the main script in src/main.py to perform real-time lane detection using the trained model, export the model to ONNX format, and visualize the results.

Training Script:

• Created src/training.py to handle model training, including data loading, loss calculation, optimization, and model checkpointing.

Visualization Utilities:

• Added utility functions in src/utils.py for visualizing input images, ground truth masks, and model predictions during training and inference.