LaneDetectionCommon // feeding LaneNet with separate images (not video)


I would like to test LaneDetection sample with various images, but I somehow cannot figure out how to do that. Is there any example how using DriveWorks SDK to read an input image and feed it into the network (CPU -> CUDA -> LaneNet -> CPU pipeline)?

The second question is about dwCameraFrameHandle_t data structure. How can I get an image from it playing a h264 video stream? I want to extract separate images out of it.


Dear alexey.abramov,

Could you please refer to page 81(Lane Detection part) and page 21(Image Pipeline) in NVIDIA DriveWorks DevGuide.pdf file? Thanks.

I have figured it out, but, sorry, your reply was useless.


I understand that it’s proprietary, but I’m curious to know where the LaneNet model/weights are stored.
The lane detection sample doesn’t point to any weight in the data folder, or make any calls to the internet to download weights, so I wonder how it gets them.