Inquiry about TensorRT Deployment for Ultra Fast Lane Detection V2

[yspark98]

Hello, my name is Yoonsoo Park, and I am a master's student at Inha University in South Korea. I am currently using Ultra Fast Lane Detection V2 for lane detection. I would like to describe my situation. I am aiming to process a file named "example.mp4" on a Jetson Orin Nano. However, due to memory issues, I perform the conversion of .pth to .onnx and .onnx to an engine file on a desktop, and then I transfer the converted engine file to the Orin Nano to run the deploy/trt_infer.py script.

Both the desktop and the Orin Nano operate under a CUDA 12.0 and TensorRT 8.6.1 environment, but when I execute deploy/trt_infer.py on the Orin Nano, I encounter a segmentation fault (core dumped) error. If there is any advice on how to successfully run UF Ldv2 on the Nvidia Jetson Orin Nano, such as matching CUDA and TensorRT versions, I would greatly appreciate your guidance.

Thank you.