How to provide input images(png/jpeg) to Lanenet Application

Autonomous Vehicles DRIVE PX 2 DriveWorks
1

Hi,

I am trying to provide stored images of type png to Lanenet, rather than using the default video / camera input.
Is there some way to give input to Lanenet or do we have to convert this png images to some format and feed Lanenet.

2

Dear puneeth.joseph,
Just for your information, Lanenet is trained with RCB images. The sample accepts only RAW/h264 video. You may need to modify the sample to fit needs. You can see in the code that DwImageCUDA is fed into lanenet handle. Here in the sample DwImageCUDA is prepared from a frame of video. You can check creating a separate dwImageCUDA buffer with your png image and get output corresponding to that image.

3

Hi Siva,

I understand that Lanenet only takes RCB images.
But I read an image from dwRenderEngine_createImageGLFromFile(png Image) and converted into DwImageCUDA.
I was able to see the image get rendered , but I don’t see any lane markings.
Is there some way we can check whether lanes are detected or not as the status is SUCCESS in all places.

4

Dear puneeth,
Can you check the output from dwLaneDetector_getLaneDetections() to know the details.

5

Siva,

It is DW_SUCCESS.

6

I am referring to check dwLaneDetection struct in dwLaneDetector_getLaneDetections().

7

The values in the structure are 0’s, which means it didn’t detect lanes.