cropped_size=o.opt.c_size,
transform=False,
input_images = [0,1,2])
#this and beat sum(120) are to use 120 long data set!
#data = torch.utils.data.Subset(data, range(120))
#beat = [8708,900,2403]
#beat = [90,10,20]
#traindata, valdata, testdata = torch.utils.data.random_split(data, beat)
octmodel = model.CapsNet(o.opt)
octmodel.to(o.opt.device)
if o.opt.loadcheckpoint:
octmodel.load_state_dict(checkpoint['model_state_dict'])
#these are only used im home testing, when randomsubset samplers are on in test and train!!!
setsize={'train':range(100),
'val':range(20)}
if o.opt.train:
sys.stdout.write('Starting Training... ' + '\n' + '\n')
sys.stdout.write('Mean ' + str(1-np.mean(self.col_losses1)) + '\n' + \
'Std ' + str(np.std(self.col_losses1)) + '\n' + \
'Min ' +str(1-np.max(self.col_losses1)) + '\n' + \
'Max ' + str(1-np.min(self.col_losses1)) + '\n')
'''
data_dir = '/media/arjun/VascLab EVO/projects/oct_ca_seg/actual final data'
#path to whichever model you want. usually will live in a ehckpoint
checkpoint = torch.load(
'/media/arjun/VascLab EVO/projects/oct_ca_seg/runsaves/Final1-pawsey/checkpoints/checkpoint.pt'
)
#checkpoint = torch.load('/group/pawsey0271/abalaji/projects/oct_ca_seg/run_saves/Final1-pawsey/checkpoints/checkpoint.pt')
model = m.CapsNet(o.opt)
model.load_state_dict(checkpoint['model_state_dict'])
model.to('cuda')
#this should be the testsamples from your loaded model
#testnames = os.listdir('/group/pawsey0271/abalaji/projects/oct_ca_seg/run_saves/Final1-pawsey/testsamples')
testnames = os.listdir(
'/media/arjun/VascLab EVO/projects/oct_ca_seg/runsaves/Final1-pawsey/testsamples'
)
#this probs wont exist yet!!!
diceorderednames = np.load(
'/media/arjun/VascLab EVO/projects/oct_ca_seg/runsaves/Final1-pawsey/analysis/diceordered.npy'
)
deploydata = DeployOCTDataset(data_dir, testnames)