Seg example - #42
Conversation
mosheraboh
left a comment
mosheraboh
left a comment
There was a problem hiding this comment.
Thanks!
Please see my comments inline.
Please also add a README file like in other examples and a short test (running just on several samples) if possible.
| from scipy.ndimage.filters import gaussian_filter | ||
| from scipy.ndimage.interpolation import map_coordinates | ||
| from torch import Tensor | ||
| import elasticdeform as ed |
There was a problem hiding this comment.
It's a new dependency - add it to requirements.txt
If there are more, please add them too.
| @@ -265,23 +274,13 @@ def aug_op_elastic_transform(aug_input: Tensor, alpha: float = 1, sigma: float = | |||
| :param channels: which channels to apply the augmentation | |||
| :return distorted image | |||
| """ | |||
There was a problem hiding this comment.
Why did you change it? the previous implementation didn't work?
I see that you are not using channels - can you support it as well?
There was a problem hiding this comment.
I find it more straightforward and easy to use.
| dx2 = gaussian_filter((random_state.rand(*shape) * 2 - 1), sigma, mode="constant", cval=0) * alpha | ||
| # convert back to torch tensor | ||
| aug_input = [numpy.array(t) for t in aug_input] | ||
| aug_input_d = ed.deform_random_grid(aug_input, sigma=7, points=3, axis=[(1, 2), (1,2)]) |
There was a problem hiding this comment.
Don't use fixed values - use the function arguments
| random crop a (3d) tensor and resize it to a given size | ||
| :param crop_size: float <= 1.0 - the fraction to crop from the original tensor for each dim | ||
| :param x_off: float <= 1.0 - the x-offset to take | ||
| :param y_off: |
|
|
||
|
|
||
| def aug_op_random_crop_and_resize(aug_input: Tensor, | ||
| out_size, |
There was a problem hiding this comment.
missing type annotation - you have more though - can you please go over the new code and add type annotations?
| # :param padding: Optional, padding size | ||
| # """ | ||
|
|
||
| # self.input_data = input_data |
There was a problem hiding this comment.
remove commented out code
| image = imread(image_fn) | ||
|
|
||
| # ====================================================================== | ||
| # TODO - change type to float if input image and to int it mask image |
| image = image.astype('float32') | ||
| # ===================================================================== | ||
|
|
||
| # img_path = self.input_data + str(inner_image_desc) + '.jpg' |
There was a problem hiding this comment.
remove commented out code and unused functions
| @@ -0,0 +1,113 @@ | |||
|
|
|||
| """ Full assembly of the parts to form the complete network """ | |||
There was a problem hiding this comment.
Where did tou copy this code from?
What is the license? Should we add the link to the original code?
| @@ -0,0 +1,584 @@ | |||
| import logging | |||
There was a problem hiding this comment.
I guess that this is old code that you should delete, right?
… of the unet code
mosheraboh
left a comment
mosheraboh
left a comment
There was a problem hiding this comment.
Thanks!
Some more minor comments.
I guess that you still plan to add readme with info, right?
What about a short unittest? can you come up with something relatively short?
|
|
||
| aug_tensor = F.interpolate(aug_tensor, out_size) | ||
|
|
||
| # else: |
| pred_name: str = None, | ||
| target_name: str = None, | ||
| filter_func: Optional[Callable]=None, | ||
| class_weights=None, |
| pred_name: str = None, | ||
| target_name: str = None, | ||
| filter_func: Optional[Callable] = None, | ||
| class_weights=None, |
| @@ -105,7 +89,7 @@ | |||
| 'force_reset_model_dir': True, # If True will reset model dir automatically - otherwise will prompt 'are you sure' message. | |||
There was a problem hiding this comment.
Let's not assume anything about ROOT and DATA_ROOT.
Please document what each one is, and ask the user to fill it in
| } | ||
|
|
||
| model = model.cuda() | ||
| # model = model.cuda() |
| ########################################## | ||
| main_out_path = '../siim_data' | ||
| dataset_path = '../siim/' | ||
| out_size_list = [256, 512] |
| out_size_list = [256, 512] | ||
|
|
||
|
|
||
| def rle2mask(rles, width, height): |
There was a problem hiding this comment.
missing type annotations and documentation in this file
|
|
||
|
|
||
|
|
||
| if __name__ == '__main__': |
There was a problem hiding this comment.
Do you require to call this script?
If yes, why isn't it part of the processor (instead of reading and saving back to the disk?_
| FuseUtilsGPU.choose_and_enable_multiple_gpus(NUM_GPUS, force_gpus=force_gpus) | ||
|
|
||
| RUNNING_MODES = ['train', 'infer', 'eval'] # Options: 'train', 'infer', 'eval' | ||
| # RUNNING_MODES = ['eval'] # Options: 'train', 'infer', 'eval' |
| @@ -0,0 +1,117 @@ | |||
| import torch | |||
There was a problem hiding this comment.
can you move it to fuse/models?
…from rle encoding
* remove fuse1 data package * remove dataset from manager * convert mnist to fuse2 style * add fuse data package * adjust mnist runner * imaging extension Co-authored-by: moshiko <moshiko.raboh@ibm.com> Co-authored-by: Alex Golts <alex.golts@ibm.com>
…o hello_world_unittest
…e nit to the user inside the notebook
Unittest for hello world notebook
3 participants
segmentation example using the eval package api