MCPcopy Create free account
hub / github.com/aim-uofa/FreeCustom / load_mask

Function load_mask

utils/utils.py:36–55  ·  view source on GitHub ↗

Load an image mask, resize it, and prepare it for use in PyTorch [0, 255] -> [0, 1], returned shape (1,1,H,W) Parameters: mask_path (str): The file path to the image mask. device (torch.device): The PyTorch device (CPU or GPU) on which the mask should be loaded. size

(mask_path, device, size=(128, 128), mode='nearest')

Source from the content-addressed store, hash-verified

34 return image
35
36def load_mask(mask_path, device, size=(128, 128), mode='nearest'):
37 """
38 Load an image mask, resize it, and prepare it for use in PyTorch [0, 255] -> [0, 1], returned shape (1,1,H,W)
39 Parameters:
40 mask_path (str): The file path to the image mask.
41 device (torch.device): The PyTorch device (CPU or GPU) on which the mask should be loaded.
42 size (tuple, optional): The target size to which the mask should be resized. Default is (64, 64).
43 mode (str, optional): The interpolation mode for resizing. Options include 'nearest', 'bilinear', 'bicubic', and more.
44 Default is 'nearest'.
45
46 Returns:
47 torch.Tensor(1,1,H,W): A PyTorch tensor with shape (1, 1, H, W) representing the resized image mask.
48
49 Example:
50 >>> mask = load_mask('mask.png', device='cuda', size=(128, 128), mode='nearest')
51 """
52 mask = read_image(mask_path)
53 mask = F.interpolate(mask.unsqueeze(0), size=size, mode=mode)
54 mask = (mask / 255.).to(torch.uint8).to(device)
55 return mask
56
57def show_cam_on_image(img: np.ndarray,
58 mask: np.ndarray,

Calls

no outgoing calls

Tested by

no test coverage detected