f_train = theano.function([input_var, target_var], [cost], updates=updates) from confusionmatrix import ConfusionMatrix batch_size = 250 num_epochs = 175 train_acc = [] valid_loss = [] valid_acc = [] train_loss = [] cur_loss = 0 loss = [] Train = DP.get_paths("/home/xvt131/Functions/Adhish_copy/Training-Rand") Test = DP.get_paths("/home/xvt131/Functions/Adhish_copy/Validating-Rand") for epoch in range(num_epochs): cur_loss = 0 val_loss = 0 confusion_valid = ConfusionMatrix(2) confusion_train = ConfusionMatrix(2) # Train = np.random.choice(Train_all, 5) for im in Train: X_train, Y_train = DP.Patch_3D_para(im, PS) num_samples_train = Y_train.shape[0] num_batches_train = num_samples_train // batch_size for i in range(num_batches_train):
f_vali = theano.function([input_var, input2_var, input3_var, target_var], [costV]) import Evaluation as E import DP1 as TD import scipy.io as io all_dice = np.array([]) with np.load( '/home/xvt131/Biomediq/Functions/Adhish_copy/Exp101/cheat.npz') as f: param_values = [f['arr_%d' % i] for i in range(len(f.files))] lasagne.layers.set_all_param_values(output, param_values) Dice = [] for img in DP.get_paths("/home/xvt131/Biomediq/Data/kneeData/Validating-Rand"): A, B, C = TD.get_indeces(img) B1 = B.reshape(np.prod(B.shape)) batch = 5000 num_batches = A.shape[0] / batch Sha = B.shape preds = np.zeros(shape=(len(B1), NC)) for i in range(num_batches): idx = range(i * batch, (i + 1) * batch) K = A[idx] M, N, O = TD.Patch_gen(K, 29, C) preds[idx] = f_eval(M, N, O) if num_batches * batch < A.shape[0]: tot = num_batches * batch
f_eval = theano.function([input_var, input2_var, input3_var, input4_var], eval_out) f_train = theano.function([input_var, input2_var, input3_var, input4_var, target_var], [cost], updates=updates) f_vali = theano.function([input_var, input2_var, input3_var,input4_var ,target_var], [costV]) import try_DP as TD import scipy.io as io import Evaluation as E with np.load("/home/xvt131/Network_adapt/triplanar_Params_WI.npz") as f: param_values = [f['arr_%d' % i] for i in range(len(f.files))] lasagne.layers.set_all_param_values(output, param_values) for img in DP.get_paths("/home/xvt131/Functions/Adhish_copy/Validating-Rand"): A, B, C = TD.Tri_Image_Load(img) B1 = B.reshape(np.prod(B.shape)) batch = 100 num_batches = A.shape[0] / batch Sha = B.shape print Sha TibiaD = [] FemoralD = [] preds = np.zeros(shape = ( len(B1), 2 )) for i in range(num_batches): idx = range(i*batch, (i+1)*batch) K = A[idx] M, N, O, P= TD.Patch_gen(K, 29, C) preds[idx] = f_eval(M,N,O, P) Final_pred = np.argmax(preds, axis = -1)
f_eval = theano.function([input_var], eval_out) f_vali = theano.function([input_var, target_var], [costV]) f_train = theano.function([input_var, target_var], [cost], updates=updates) import Evaluation as E import try_DP as TD import scipy.io as io with np.load("/home/xvt131/Functions/Adhish_copy/3D_params/3D_all_params11.npz" ) as f: param_values = [f['arr_%d' % i] for i in range(len(f.files))] lasagne.layers.set_all_param_values(output, param_values) Dice = [] for img in DP.get_paths("/home/xvt131/Functions/Adhish_copy/vseries"): print img A, B, C = TD.Tri_Image_Load(img) B1 = B.reshape(np.prod(B.shape)) batch = 5000 num_batches = A.shape[0] / batch Sha = B.shape preds = np.zeros(shape=(len(B1), 2)) for i in range(num_batches): idx = range(i * batch, (i + 1) * batch) K = A[idx] M = TD.Patch_gen_three(K, PS, C) M = M.reshape(batch, 1, PS, PS, PS) preds[idx] = f_eval(M)
f_train = theano.function([input_var, input2_var, input3_var, target_var], [cost], updates=updates) from confusionmatrix import ConfusionMatrix batch_size = 15 num_epochs = 50 train_acc= [] train_loss = np.array([]) valid_acc, valid_loss = [], [] test_acc, test_loss = [], [] cur_loss = 0 loss = [] Train = DP.get_paths("/home/xvt131/Functions/Adhish_copy/Training") import gc for epoch in range(num_epochs): cur_loss = 0 #Train = np.random.choice(Train_all, 5) confusion_valid = ConfusionMatrix(3) confusion_train = ConfusionMatrix(3) for im in Train: XY, XZ, YZ, Y_train = DP.Patch_triplanar_para(im, PS) num_samples_train = Y_train.shape[0] num_batches_train = num_samples_train // batch_size for i in range(num_batches_train): idx = range(i*batch_size, (i+1)*batch_size)
f_vali = theano.function([input_var, input2_var, input3_var, target_var], [costV]) from confusionmatrix import ConfusionMatrix batch_size = 250 num_epochs = 175 train_acc = [] valid_acc = [] cur_loss = 0 loss = [] valid_loss = [] Train = DP.get_paths("/home/xvt131/Functions/Adhish_copy/croppedT") Test = DP.get_paths("/home/xvt131/Functions/Adhish_copy/croppedV") import gc for epoch in range(num_epochs): cur_loss = 0 val_loss = 0 confusion_valid = ConfusionMatrix(2) confusion_train = ConfusionMatrix(2) for im in Train: XY, XZ, YZ, Y_train = DP.Patch_triplanar_para(im, PS) num_samples_train = Y_train.shape[0] num_batches_train = num_samples_train // batch_size for i in range(num_batches_train): idx = range(i * batch_size, (i + 1) * batch_size)