Hi,
I'm trying to create an algorithm for lane detection/lane marking segmentation that is based on a deep learning model on the TDA2P EVM. I want to use the TIDL tools for converting my model and then integrate it into an existing PSDK use case (like tidl_OD). Unfortunately, the model I chose (https://github.com/MaybeShewill-CV/lanenet-lane-detection) is not compatible with TIDL (unsupported topology/operators) and all other validated TIDL models are object detection or image classification.
Do you know some lane detection models that are compatible with TIDL?
Do you have any tips how I could implement such a model? I was thinking about using a semantic segmentation model that can detect lane markings and then process the output of that. But for that I would again need a SemSeg model that is compatible with TIDL...
Regards,
Tobias