search

IrohXu / lanenet-lane-detection-pytorch

Unofficial implemention of lanenet model for real time lane detection Pytorch Version
MIT License
150 stars 39 forks source link

How to use pre-trained model and where is post-processing codes? #5

Closed landkwon94 closed 2 years ago

landkwon94 commented 2 years ago

Hello, Sir. First of all, many thanks for your nice codes!

Because there were so many errors while using original-LaneNet code, I was looking for other version of same network. And I found your codes, and I saw it works well!

By the way, I have some questions.

1) I want to use pre-trained model (such as trained LaneNet model which was trained with tuSimple datasets), may I ask you direction how to use it in your codes?

2) I was to use post processing codes. Because I want to count the number of lanes from the image. But I cannot find this code... Where can I revise this codes?

3) Also, I want to classify multi-class lane detection.. Is there any config file in your codes? And do I have to revise some parts of the code for multi-class detection?

Thank you again for your codes! I will wait for your reply :)

IrohXu commented 2 years ago

Hi, landkwon94:

  1. You can use my pre-trained model here: https://github.com/IrohXu/lanenet-lane-detection-pytorch/blob/main/log/best_model.pth The model is ENet version and trained with Tusimple dataset. For using the other model trained by yourself, please change the --model of test.py.

  2. You can add your code inside the test.py.

  3. Tusimple will not helpful for multi-class lane detection & classification, as it do not label the type of the lane marking. My group is researching for this topic these days and we will submit a new dataset and new transformer-based lane detection model to ECCV.

landkwon94 commented 2 years ago

Thank you so much for your kind reply! I will carefully do that, and I will wait for your new paper! Thanks again :)