search

ApolloAuto / apollo

An open autonomous driving platform
Apache License 2.0
25.01k stars 9.67k forks source link

About Lane Line Detection & Post-Processing #11074

Open mertmerci opened 4 years ago

mertmerci commented 4 years ago

I’m trying to understand the perception module and confused on lane line detection/post-processing steps. First of all, is there any post-processing after lane-line detection and what is the output of the process? Is the output a segmented picture?

I guess, after the CNN implementation for lane line detection (pathway: modules/perception/production/data/perception/camera/models/lane_detector/darkSCNN/dark_SCNN_vpt_8x_trainable_xavier_lighter_ud_deploy.pt) a segmented picture is obtained with different lines, etc. but in order to work with these lines, they should be translated into functions or coordinates and these should be made under a post-processing section where we can work for thresholding.

Is my guess correct? If yes, what is the location of this file?

markofan commented 4 years ago

Do you understand the Post-Processing step? Is there any thresholding? Do you understand the process2D function?

mertmerci commented 4 years ago

Well, actually not on a satisfied level. I'm still waiting for the response of this issue.

markofan commented 4 years ago

I want to know that "Lane detection" and "object detection" use different models, which are darkscnn and yolo, is this right?

ashwinsarvesh commented 4 years ago

can someone answer this question?