I’m new to pytorch.I override the getitem() in Dataset:
ct_path= self.ct_list[index] seg_ path =self.seg_list[index] ct_array = sitk.GetArrayFromImage(ct) seg_array = sitk.GetArrayFromImage(seg)
the shape of ct_array usually is 200 * 256 * 256, I wanna every ct_array is divided into 20 parts, and every parts is 10 * 256 *256， because my network only can run with this size.
As I understand，with one ‘index’, I can get one sample.So I must return a big tuple of 20 * 10 * 256 *256.But in my training, I hope my batch_size is 6, and take a 10 * 256 *256 tuple as a sample. If I define the batch_size=6 in DataLoader, it will return 6 tuple of 200 * 256 * 256, and these six 200 * 256 *256 tuple will be train together, what I want is six 10 * 256 *256 tuple.How can I do it?