Self-Driving Car Using Behavioral Cloning (Graduation Project)

Graduation project training CNNs on manual driving data to replicate human driving behavior.

End-to-end learning approach that maps images to driving commands using CNNs.

Demonstrate feasibility of behavioral cloning for autonomous driving in controlled settings.

Collected/curated data, engineered augmentations, designed/trained CNN models, tuned with GPU, and validated performance.

Data pipeline → CNN training → evaluation; augmentation for balance; dropout and tuning to generalize.

Mitigated dataset imbalance with augmentation; reduced overfitting via regularization and careful splits.

Delivered a crash-free autonomous run in a controlled environment and built strong ML/CV foundations.

Expand datasets, test diverse conditions, and explore RL for adaptive behavior.