def _cam_image(
self, norm_image: torch.Tensor, class_idx: Optional[int] = None
) -> Dict[str, Union[np.ndarray, Dict[str, np.ndarray]]]:
"""Get CAM for an image.
Args:
norm_image (torch.Tensor): Normalized image.
class_idx (:obj:`int`, optional): Class index for calculating GradCAM.
If not specified, the class index that makes the highest model
prediction score will be used.
Returns:
Dictionary containing unnormalized image, heatmap and CAM result.
"""
image = unnormalize(norm_image, self.mean, self.std) # Unnormalized image
norm_image_cuda = norm_image.clone().unsqueeze_(0).to(self.device)
heatmap, result = {}, {}
for layer, gc in self.gradcam.items():
mask, _ = gc(norm_image_cuda, class_idx=class_idx)
cam_heatmap, cam_result = visualize_cam(
mask,
image.clone().unsqueeze_(0).to(self.device)
)
heatmap[layer], result[layer] = to_numpy(cam_heatmap), to_numpy(cam_result)
return {
'image': to_numpy(image),
'heatmap': heatmap,
'result': result
}
def unnormalize(self, image, transpose=False):
"""Un-normalize a given image.
Args:
image (numpy.ndarray or torch.Tensor): A ndarray
or tensor. If tensor, it should be in CPU.
transpose (bool, optional): If True, transposed output will
be returned. This param is effective only when image is
a tensor. If tensor, the output will have channel number
as the last dim. (default: False)
"""
return unnormalize(image, self.mean, self.std, transpose)
def unnormalize(self, image, transpose=False, data_type=None):
"""Un-normalize a given image.
Args:
image (numpy.ndarray or torch.Tensor): A ndarray
or tensor. If tensor, it should be in CPU.
transpose (bool, optional): If True, transposed output will
be returned. This param is effective only when image is
a tensor. If tensor, the output will have channel number
as the last dim. (default: False)
data_type (str, optional): Type of image. Required only when
dataset has multiple types of images. (default: None)
"""
mean = self.mean if data_type is None else self.mean[data_type]
std = self.std if data_type is None else self.std[data_type]
return unnormalize(image, mean, std, transpose)
def _cam_image(self, norm_image, class_idx=None):
"""Get CAM for an image.
Args:
norm_image: Normalized image. Should be of type
torch.Tensor
Returns:
Dictionary containing unnormalized image, heatmap and CAM result.
"""
image = unnormalize(norm_image, self.mean,
self.std) # Unnormalized image
norm_image_cuda = norm_image.clone().unsqueeze_(0).to(self.device)
heatmap, result = {}, {}
for layer, gc in self.gradcam.items():
mask, _ = gc(norm_image_cuda, class_idx=class_idx)
cam_heatmap, cam_result = visualize_cam(
mask,
image.clone().unsqueeze_(0).to(self.device))
heatmap[layer], result[layer] = to_numpy(cam_heatmap), to_numpy(
cam_result)
return {'image': to_numpy(image), 'heatmap': heatmap, 'result': result}