Testing the model on another image type

[darkdevahm]

I would like to test your model on another image that have different resolution (attached below) I would like were can I find the parameters in the post process to adjust for it to view the lane markings on this image.

EDIT:: I have been through other issues in this repo , and I found that I need to edit the post process , I'm a quite new to deep learning and CV , it would be great if someone can advise me on what parts exactly to look on and change in the post process

PS. I have resized the image to suit the resolution of the TuSimple test images , the model runs , but it views the src image the same as the image I inserted and the detected lanes are not added, and the masked image shows a black image.

[MaybeShewill-CV]

@an-ahmed You need a new remap.yml which is used to convert the source image from front view into bird view. You can read the post process module for details:)

[darkdevahm]

Thanks for your swift reply and time, I have seen another closed issue linking to on of your repo on how to generate remap.yml files :)

My qestion regarding the remap.yml file , do I need to make it for all test images from the same distribution , or making it for only one image will be enough?

[MaybeShewill-CV]

@an-ahmed One is enough. The remap.yml for Tusimple dataset was generated via one of the testing image in the dataset:)

[darkdevahm]

Thanks alot :)

[MankaranSingh]

@MaybeShewill-CV Can you point to a good way of making a custom remap.yml file ?
A tutorial/simple description would really help. Once I am successful for generating ipm file for my dataset, iil add a pull request of full steps to do so.

[MaybeShewill-CV]

@MankaranSingh 1.For ipm image you may simply do a perspective transformation to get it. You may google how to do it. 2.Actually you can do your own post process procedure instead of using my version here:)