can not train of the dataset i set

[hyp7266]

why it can train of culane ，but can not train of a small data set builded with the same format

[hyp7266]

Do you have any requirements on the picture during the training of this code? Other codes of lane lines can be used

[Wolfwjs]

Hi, we have no special requirements for images, the annotation file annotates the (x, y) coordinates of each line. Can you describe your problem in more detail?

[lzyuan168]

Hi @Wolfwjs I have my custom dataset in this format and structure. May I know if it is the correct and required format for training.

dataset_name |---sub_folder_1 |---000.jpg |---000.lines.txt . . . |---2000.jpg |---2000.lines.txt

|---sub_folder_2 |---000.jpg |---000.lines.txt . . . |---2000.jpg |---2000.lines.txt

|---list |---train.txt |---val.txt |---test.txt

May I know if it is necessary to have the segmentation ground truth text / image files in order for the training to run?

Thank you.