ROI_img = ori_img[BeginY:, ally[0]:ally[1], :].copy()
ROIShape = ROI_img.shape
FullImage = np.zeros(imshape)
FinalShape = FinalShape * 1
pred_imgFinal = cv2.resize(FinalShape.astype(np.uint8), (ROIShape[1], ROIShape[0]))
FullImage[BeginY:, ally[0]:ally[1]] = pred_imgFinal
NucluesImage = MatchTemplate(FullImage)
NucluesImage = NucluesImage[BeginY:, ally[0]:ally[1]]
tmp_img = np.zeros(imshape)
tmp_label = FullImage.copy()
# with open('./logs/data/%s_side.pkl'%img_path,'w+') as f:
# data = [tmp_img, tmp_label, img_path, ally]
# pkl.dump(data, f)
MSE_loss = MSE_pixel_loss(tmp_img, tmp_label, img_path, ally)
print 'sideoutput:', img_path, MSE_loss
MSE_loss_list.append(MSE_loss)
if save_test_img:
pred_img = cv2.resize(new_Image.astype(np.uint8), (ROI_img.shape[1], ROI_img.shape[0]))
tmp_img = pred_img.copy()
tmp_ROI_img = ROI_img.copy()
pred_img[pred_img==3]=2
pred_img[NucluesImage==1]=3
ROI_img = crop_boundry(ROI_img, pred_img)
alpha_img = ori_img.copy()
alpha_img[BeginY:, ally[0]:ally[1]] = ROI_img
ROI_img = cv2.addWeighted(alpha_img, 0.4, ori_img, 0.6, 0)
## imshow the truth
## imshow the truth
ROI_img = get_truth(ROI_img, img_path, ally, BeginY)
if not os.path.exists(os.path.join(args.results, model_name)):
os.mkdir(os.path.join(args.results, model_name))
save_name = os.path.join(args.results, model_name, img_path)
cv2.imwrite(save_name, ROI_img)
# predict the boundary
FullImage = np.zeros(imshape)
FinalShape = FinalShape * 1
pred_imgFinal = cv2.resize(FinalShape.astype(np.uint8),
(ROIShape[1], ROIShape[0]))
FullImage[BeginY:, ally[0]:ally[1]] = pred_imgFinal
# cv2.imwrite('./tmp_data/%s.png' % img_path.split('.')[0], FullImage)
tmp_img = np.zeros(imshape)
MSE_loss = MSE_pixel_loss(tmp_img, FullImage, img_path, ally)
MSE_loss_list.append(MSE_loss)
MSE_loss = np.mean(np.stack(MSE_loss_list), 0)
if np.mean(MSE_loss) < mean_loss:
torch.save(model.state_dict(),
'./models/%s/best.pth' % (model_name))
print 'success save the best model'
mean_loss = np.mean(MSE_loss)
end = time.time()
# print('epoch: {:d} | lr: {:f} | MSE_loss: {:s}'.format(epoch, args.lr, MSE_loss))
print('epoch: {:d} | lr: {:f} | time: {:.2f}| MSE_loss: {:s}'.format(
epoch, args.lr, end - start, MSE_loss))
with open('./logs/test_%s_log.txt' % (model_name), 'a+') as f:
log_str = ('epoch: {:d} | lr: {:f} | MSE_loss: {:s}'.format(
epoch, args.lr, MSE_loss))
f.writelines(log_str + '\n')