膨胀卷积后的mask数据集有吗

[HarryHsing]

膨胀卷积后的mask数据集有吗

[Jasonshao0123]

请问亲找到mask50的数据了？

[HarryHsing]

请问亲找到mask50的数据了？

This is my code of implementation, hope it helps

`
import os
import glob
from skimage import io
from skimage.morphology import dilation, disk

def apply_dilation_to_masks(input_dir, output_dir, struct_elem_size=50):
if not os.path.exists(output_dir):
os.makedirs(output_dir)

mask_files = glob.glob(os.path.join(input_dir, '*.*'))  # Adjust to read all image files
structuring_element = disk(struct_elem_size)

for mask_file in mask_files:
    # Load mask in grayscale
    mask = io.imread(mask_file, as_gray=True)
    mask = (mask > 0).astype('uint8')  # Ensure binary mask (0 or 1)
    mask_dilated = dilation(mask, structuring_element)  # Apply dilation
    mask_dilated = (mask_dilated * 255).astype('uint8')  # Convert to 0-255 range
    
    # Save dilated mask as a .png file
    output_path = os.path.join(output_dir, os.path.basename(mask_file))
    output_path = os.path.splitext(output_path)[0] + '.png'
    io.imsave(output_path, mask_dilated)  # Save dilated mask

input_directory = './SRD/train/train_B' # Replace with the path to your original masks
output_directory = './SRD/train/train_B_dilated' # Replace with the path to save the dilated masks

apply_dilation_to_masks(input_directory, output_directory)
`

[Jasonshao0123]

谢谢回复，后来仔细看了原作者的论文，也可以使用引用的论文“From Shadow Generation to Shadow Removal”中的matlab代码获得mask50数据集

[HarryHsing]

谢谢回复，后来仔细看了原作者的论文，也可以使用引用的论文“From Shadow Generation to Shadow Removal”中的matlab代码获得mask50数据集

This is the python version of that matlab code implemented by meself.

[Jasonshao0123]

请问亲试过将batch_size设置为2 甚至更大吗？为什么设置为更大训练的时候会报错呢？图片显示多了一个维度