"Camera-only self driving vehicle dataset for end to end trajectory prediction and vehicle control in autonomous driving"

"Autonomous driving constitutes a rapidly advancing field that leverages artificial intelligence, computer vision, and machine learning to enable safe and efficient vehicle navigation. However the barrier to enter this field of research and product development is high, requiring expensive sensor arrays and proprietary, large scale datasets. This work explores a camera-only approach to autonomous navigation for passenger vehicles, aiming to reduce both system complexity and implementation costs. The proposed methodology relies on simulated environments for data collection, model training, and evaluation, allowing for rapid iteration and testing under diverse conditions. Three heterogeneous deep learning architectures\u2014namely Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Transformers\u2014were investigated for their ability to process visual input and predict future vehicle trajectories. The findings suggest that the camera-only end-to-end approach, combined with simulation-based development, holds potential as a scalable and cost-effective foundation for future autonomous systems. This dataset represents the data collected and used to produce these findings."