search

xuxuxuxuxuxjh / LB-UNet

LB-UNet: A Lightweight Boundary-assisted UNet for Skin Lesion Segmentation
8 stars 0 forks source link

How to create Prompt? #1

Open saisusmitha opened 3 hours ago

saisusmitha commented 3 hours ago

In dataset.py- you have mentioned points_list = os.listdir(path_Data+'train/points_boundary2/')- where are these available? how to create them? Does boundary.py create these? In boundary.py there is no creation of such files. I am confused kindly let me know how to create prompts.

xuxuxuxuxuxjh commented 3 hours ago

Yes, you can use boundary.py to create boundary maps. And you can place these maps wherever, as long as they correspond to the directory names in dataset.py.

saisusmitha commented 45 minutes ago

Thanks for replying, I already did that 1) I ran your code exactly and best_val is being printed but no png is being saved. So I thought there is no line creating the output_path2 directory in the code

2) So I just added a line "Path(output_path2).mkdir(parents=True, exist_ok=True)" to create the folder- then images were saved into the output_path2 folder - they were taking a lot of time but images are being saved - they look empty - completely black and around 650-670 bytes that's it.

Can you please tell me why this is happening and show me a sample image saved by boundary.py showing what it looks like? what exactly the boundary.py is saving?

ISIC_0000024_segmentation ISIC_0000024_segmentation

xuxuxuxuxuxjh commented 24 minutes ago

image image

xuxuxuxuxuxjh commented 5 minutes ago

Just modify the data path, and then the model will run well. But if your data images are not skin lesion images, maybe having more than two connected blocks, the model will fail.