Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

normalizing frames with custom mean_std instead of defaulting to resnet #316

Closed
wants to merge 2 commits into from
Closed

normalizing frames with custom mean_std instead of defaulting to resnet #316

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2c0db74
passing custom mean_std at instanciation
Jan 12, 2023
a43235e
brought back norm_resnet()
Jan 12, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 22 additions & 2 deletions aloscene/frame.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -465,6 +465,9 @@ def norm_minmax_sym(self):
tensor = 2 * tensor - 1.0
elif tensor.normalization == "255":
tensor = 2 * (tensor / 255.0) - 1.0
elif tensor.mean_std is not None:
tensor = tensor.norm01()
tensor = 2 * tensor - 1.0
else:
raise Exception(f"Can't convert from {tensor.normalization} to norm255")
tensor.mean_std = None
Expand All @@ -484,7 +487,7 @@ def mean_std_norm(self, mean, std, name) -> Frame:
"""
tensor = self
mean_tensor, std_tensor = self._get_mean_std_tensor(
tensor.shape, tensor.names, tensor._resnet_mean_std, device=tensor.device
tensor.shape, tensor.names, (mean, std), device=tensor.device
)
if tensor.normalization == "01":
tensor = tensor - mean_tensor
Expand All @@ -507,10 +510,27 @@ def mean_std_norm(self, mean, std, name) -> Frame:

return tensor

def norm_meanstd(self, mean_std=None, name=None) -> Frame:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Having two methodes with nearly the same name is not a good practice. try to fix mean_std_norm instead

"""Returns z-norm of the current frame.
This method will simply call `frame.mean_std_norm()` with the mean/std property of the frame and the selected name.
Instead of a custom mean/std, you can use the resnet norm based on the normalized use of resnet on pytorch.
Examples
--------
>>> frame_resnet = frame.norm_resnet(name="resnet")
>>> frame_custom_norm = frame.norm_resnet(name="custom")
"""
if name == "resnet":
return self.norm_resnet()
elif mean_std is not None and len(mean_std) == 2:
return self.mean_std_norm(mean=mean_std[0], std=mean_std[1], name=name)
elif self.mean_std is not None:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

there's no difference between this condition and the one above.

return self.mean_std_norm(mean=self.mean_std[0], std=self.mean_std[1], name=name)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The attribute mean_std is used to store the current normalization parameters. You can delete this

else:
raise Exception("Please pass a mean_std tuple or use the resnet norm")

def norm_resnet(self) -> Frame:
"""Normalized the current frame based on the normalized use on resnet on pytorch. This method will
simply call `frame.mean_std_norm()` with the resnet mean/std and the name `resnet`.

Examples
--------
>>> frame_resnet = frame.norm_resnet()
Expand Down