2d segmentation with inputs of .png files about niftynet HOT 2 CLOSED

Hi @mariamhrr , at the moment we don't have any concrete example of 2d image segmentation and png format is not supported. Contributions are welcomed!

from niftynet.

@mariamhrr @wyli Since the merge of image-loader NiftyNet now fully supports 2D images without having to convert them to NiftyFormat. There is no need to fake 3-dimensional spatial_window_size either (although it is perfectly ok if you do).

NiftyNet currently supports loading images from OpenCV, Scikit-Image and PIL/Pillow, assuming that you have the corresponding libraries installed. You can specify a loader in the config file, or it will automatically try to find one that can handle yout images.

Pretty much every 2D image format is supported (as long as it can be loaded by one of those libraries): .jpg, .png, .bpm, .tif ...

For more information check image_loader.py.

Example configuration file:

[images]
path_to_search = ~/data/test_image_segmentation/
filename_contains = bmp
spatial_window_size = (224, 224)
interp_order = 1
loader = opencv

[masks]
path_to_search = ~/data/test_image_segmentation/
filename_contains = _mask
spatial_window_size = (224, 224)
interp_order = 0
loader = skimage

from niftynet.