get the input images in the files of spine_localization_dataset

[BruceFeiFei]

Hello，can you give me the input images in the files of spine_localization_dataset

[christianpayer]

Hi, the dataset is from the CSI 2014 challenge (http://csi-workshop.weebly.com/challenges.html). It seems that the links on the website to the data are not working anymore. I got an updated link from Ben Glocker, who was one of the challenge organizers: https://imperialcollegelondon.app.box.com/s/erhcm28aablpy1725lt93xh6pk31ply1 There you should find the training and testing data.

[BruceFeiFei]

Hi, the dataset is from the CSI 2014 challenge (http://csi-workshop.weebly.com/challenges.html). It seems that the links on the website to the data are not working anymore. I got an updated link from Ben Glocker, who was one of the challenge organizers: https://imperialcollegelondon.app.box.com/s/erhcm28aablpy1725lt93xh6pk31ply1 There you should find the training and testing data.

Thank you for your link！I have already got the model that trained from the 'cep' files. I still want to ask that how can I test my data by using this model.

[christianpayer]

The scripts for the cep dataset should perform validation/testing after specific step intervals during training and after the last iteration. So if you just ran the scripts, the cross validation and/or testing outputs should have been written already.

In the framework, you can load previously trained models by specifying the parameter load_model_filename of MainLoop. E.g., you can set something like this in the last line of __init__ in main.py: self.load_model_filename = 'output/weights/model-40000.index'. Adapt the path accordingly to your setup. If you would just run the testing, replace in the same file loop.run() with loop.run_test(), which will not do training, but only testing. Make sure to use the correct weight files for testing, as the framework will not check if the loaded weight file and current cross validation index correspond.

[BruceFeiFei]

The scripts for the cep dataset should perform validation/testing after specific step intervals during training and after the last iteration. So if you just ran the scripts, the cross validation and/or testing outputs should have been written already.

In the framework, you can load previously trained models by specifying the parameter load_model_filename of MainLoop. E.g., you can set something like this in the last line of __init__ in main.py: self.load_model_filename = 'output/weights/model-40000.index'. Adapt the path accordingly to your setup. If you would just run the testing, replace in the same file loop.run() with loop.run_test(), which will not do training, but only testing. Make sure to use the correct weight files for testing, as the framework will not check if the loaded weight file and current cross validation index correspond.

OK, I understand. I still have some questions about the images for testing. I have already performed the file 'convert_bmp_to_nii.py' to transform my images. And then I have already changed the parameter 'original_image_extend' in 'main.py' and got the model. How can I use it to test my dataset?