def evaluate_epoch(self):
"""
evaluating in a epoch
:return:
"""
self.eval_losses = utils.AverageMeter()
self.eval_top1 = utils.AverageMeter()
self.eval_top5 = utils.AverageMeter()
# Set the model to be in testing mode (for dropout and batchnorm)
self.model.net.eval()
for batch_idx, (batch_x, batch_y) in enumerate(self.val_loader):
if torch.cuda.is_available():
batch_x, batch_y = batch_x.cuda(
async=self.config['async_loading']), batch_y.cuda(
async=self.config['async_loading'])
batch_x_var, batch_y_var = Variable(batch_x), Variable(batch_y)
self.evaluate_step(batch_x_var, batch_y_var)
utils.view_bar(batch_idx + 1, len(self.val_loader))
def load_gt_rois(self):
# load the ground truth regions of image set
cache_file = os.path.join(self.cache_path, self.name + '_gt_roidb.pkl')
if os.path.exists(cache_file):
print('Load gt roidb from {}'.format(cache_file))
with open(cache_file, 'rb') as f:
gt_roidb = pickle.load(f)
print('Load finished')
return gt_roidb
gt_roidb = []
for i in range(len(self.image_index)):
gt_roi = self.load_gt_roi(self.image_index[i])
gt_roidb.append(gt_roi)
utils.view_bar(
'Load gt roidb from {}'.format(self.image_index[i] + '.xml'),
i, len(self.image_index))
with open(cache_file, 'wb') as f:
pickle.dump(gt_roidb, f, pickle.HIGHEST_PROTOCOL)
print('Write gt roidb to {}'.format(cache_file))
return gt_roidb
def evaluate_accuracy(self, data_iterator, net):
"""
compute top-1 accuracy
:param data_iterator:
:param net:
:return:
"""
loss = utils.AverageMeter()
acc = mx.metric.Accuracy()
for idx, (d, l) in enumerate(data_iterator):
data = d.as_in_context(self.ctx[0])
label = l.as_in_context(self.ctx[0])
output = net(data)
_loss = self.get_loss(output, label)
curr_loss = nd.mean(_loss).asscalar()
loss.update(curr_loss, data.shape[0])
predictions = nd.argmax(output, axis=1)
acc.update(preds=predictions, labels=label)
utils.view_bar(idx + 1, len(data_iterator)) # view_bar
return acc.get()[1], loss.avg
bic_s0[i] = S0_BIC
bic_dolp[i] = DoLP_BIC
bic_aop[i] = AoP_BIC
# Calculate the PSNR of S0, DoLP and AoP obtained through PDCNN method
total_S0_PSNR[i]= psnr(S0_true, S0_hat_test, 2)
total_DoLP_PSNR[i] = psnr(DoLP_true, DoLP_hat_test, 1)
total_AoP_PSNR[i] = psnr(AoP_true, AoP_hat_test, math.pi/2.)
# Calculate the PSNR of S0, DoLP and AoP obtained through BICUBIC method
total_S0_PSNR_BIC[i] = psnr(S0_true, S0_BIC, 2)
total_DoLP_PSNR_BIC[i] = psnr(DoLP_true, DoLP_BIC, 1)
total_AoP_PSNR_BIC[i] = psnr(AoP_true, AoP_BIC, math.pi/2.)
# show the progress bar
view_bar(i, IMG_NUM)
print('\n========================================Testing=======================================' +
'\n ————————————————————————————————————————————————————————————————————————————————' +
'\n| PSNR of S_0 using SRCNN: %.5f | PSNR of S_0 using BICUBIC: %.5f |' % (np.mean(total_S0_PSNR), np.mean(total_S0_PSNR_BIC)) +
'\n| PSNR of DoLP using PDCNN: %.5f | PSNR of DoLP using BICUBIC: %.5f |' % (np.mean(total_DoLP_PSNR), np.mean(total_DoLP_PSNR_BIC)) +
'\n| PSNR of AoP using SRCNN: %.5f | PSNR of AoP using BICUBIC: %.5f |' % (np.mean(total_AoP_PSNR), np.mean(total_AoP_PSNR_BIC)) +
'\n ————————————————————————————————————————————————————————————————————————————————')
print('\nSRCNN time: {} sec'.format(total_time / IMG_NUM))
for j in output_images:
imgio.imsave("./images/bic_s0_{}_{}.jpg".format(j, total_S0_PSNR_BIC[j-1]), bic_s0[j-1, 390:-470, 690:-490])
imgio.imsave("./images/pred_s0_3_path_{}_{}.jpg".format(j, total_S0_PSNR[j-1]), pred_s0[j-1, 200:-80, 325:-275])
imgio.imsave("./images/org_s0_{}.jpg".format(j), origin_s0[j-1, 390:-470, 690:-490])
fig = plt.gcf()
height, width = AoP_hat_test.shape
fig.set_size_inches(width/300, height/300)
plt.gca().xaxis.set_major_locator(plt.NullLocator())
plt.gca().yaxis.set_major_locator(plt.NullLocator())
plt.subplots_adjust(top=1, bottom=0, left=0, right=1, hspace=0, wspace=0)
plt.margins(0, 0)
plt.imshow(AoP_hat_test, cmap=cm.jet)
hm = fig2array(fig)
S0_hat_test = np.uint8(normalize(S0_hat_test, 0, 255))
DoLP_hat_test = np.uint8(normalize(DoLP_hat_test, 0, 255))
S0_hat_test = np.stack([S0_hat_test] *3 , axis=-1)
DoLP_hat_test = np.stack([DoLP_hat_test] * 3, axis=-1)
s0_writer.write(S0_hat_test)
dolp_writer.write(DoLP_hat_test)
view_bar(cur_frame-1, total_frame)
cur_frame += 1
test_video.release()
s0_writer.release()
dolp_writer.release()
cv2.destroyAllWindows()