Researchers have optimized a learning-based lane detection algorithm for the NVIDIA Jetson Nano, achieving processing speeds about 22 times faster than a previous model by converting it to TensorRT.