plt.figure(figsize=(6, 12)) n_images = np.min((seg_pred.shape[0], max_n_images)) for i in range(n_images): seg_pred[i][0, :6] = np.array([0,1,2,3,4,5]) seg[i,0,0,:6] = np.array([0,1,2,3,4,5]) plt.subplot(n_images, 2, 2*i+1) plt.imshow(seg[i, 0]) plt.subplot(n_images, 2, 2*i+2) plt.imshow(seg_pred[i]) plt.savefig(os.path.join(results_dir, "some_segmentations_ep_%d.png"%epoch)) while epoch < 50: data_gen_train = SegmentationBatchGeneratorBraTS2014(all_patients, BATCH_SIZE, validation_patients, PATCH_SIZE=(180, 164), mode="train", ignore=[81], losses=losses, num_batches=1500, seed=None) data_gen_train = seg_channel_selection_generator(data_gen_train, [2]) data_gen_train = rotation_generator(data_gen_train) data_gen_train = elastric_transform_generator(data_gen_train, 200., 14.) data_gen_train = pad_generator(data_gen_train, PATCH_SIZE) data_gen_train = center_crop_seg_generator(data_gen_train, (180, 164)) data_gen_train = Multithreaded_Generator(data_gen_train, 8, 50) data_gen_train._start() print "epoch: ", epoch train_loss = 0 train_acc_tmp = 0 train_loss_tmp = 0 batch_ctr = 0 for data, seg, idx in data_gen_train: if batch_ctr != 0 and batch_ctr % int(np.floor(n_batches_per_epoch/n_feedbacks_per_epoch)) == 0: print "number of batches: ", batch_ctr, "/", n_batches_per_epoch print "training_loss since last update: ", train_loss_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch), " train accuracy: ", train_acc_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch) all_training_losses.append(train_loss_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch))
# computes a moving average on the losses avg_loss = np.mean(losses[:, 1]) loss_new = (loss_old + losses/avg_loss) / 2. return loss_new''' losses = np.ones(len(memmap_gt)) def update_losses(losses, idx, loss): losses[idx] = (losses[idx] + loss*2.) / 3. return losses n_epochs = 40 auc_scores=None for epoch in range(0,n_epochs): data_gen_train = memmapGenerator_allInOne_segmentation_lossSampling(memmap_data, memmap_gt, BATCH_SIZE, validation_patients, mode="train", ignore=[40], losses=losses) data_gen_train = seg_channel_selection_generator(data_gen_train, [2]) data_gen_train = rotation_generator(data_gen_train) data_gen_train = center_crop_generator(data_gen_train, (PATCH_SIZE, PATCH_SIZE)) data_gen_train = elastric_transform_generator(data_gen_train, 550., 20.) data_gen_train = Multithreaded_Generator(data_gen_train, 12, 100) data_gen_train._start() print "epoch: ", epoch train_loss = 0 train_acc_tmp = 0 train_loss_tmp = 0 batch_ctr = 0 for data, seg, idx in data_gen_train: if batch_ctr != 0 and batch_ctr % int(np.floor(n_batches_per_epoch/n_feedbacks_per_epoch)) == 0: print "number of batches: ", batch_ctr, "/", n_batches_per_epoch print "training_loss since last update: ", train_loss_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch), " train accuracy: ", train_acc_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch) all_training_losses.append(train_loss_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch)) all_training_accuracies.append(train_acc_tmp/np.floor(n_batches_per_epoch/n_feedbacks_per_epoch))