images = np.zeros((1, IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS), dtype=np.uint8)
test_image = plt.imread(
"D:\\Projects\\mgr\\test\\4_test.jpg")[:, :, :IMG_CHANNELS]
test_image = resize(test_image, (IMG_HEIGHT, IMG_WIDTH), preserve_range=True)
images[0] = test_image
preds_test = model.predict(images, verbose=1)
label = labels[0]
pred_label = preds_test[0]
f, axarr = plt.subplots(2, 3)
axarr[1][2].imshow(images[0])
axarr[1][2].set_title("original")
axarr[1][1].imshow(heatmap_to_rgb(label, id2code))
axarr[1][1].set_title("truth")
axarr[1][0].imshow(heatmap_to_rgb(pred_label, id2code))
axarr[1][0].set_title("prediction mask")
for i in range(label.shape[2]):
if i > 3:
axarr[1][i - 3].imshow(pred_label[:, :, i])
axarr[1][i - 3].set_title("layer " + id2code[i + 1])
else:
axarr[0][i].imshow(pred_label[:, :, i])
axarr[0][i].set_title("layer " + id2code[i + 1])
plt.show()
batch_size=18,
epochs=600,
callbacks=callbacks)
model.save('saved_models/' + model_name + '.model')
y_pred = model.predict(test_inputs)
y_predi = y_pred
for i in range(TEST_LENGTH):
f, axarr = plt.subplots(2, 5)
axarr[1][4].imshow(test_inputs[i])
axarr[1][4].set_title("original")
axarr[1][3].imshow(heatmap_to_rgb(test_labels[i], id2code))
axarr[1][3].set_title("truth")
axarr[1][2].imshow(heatmap_to_rgb(y_pred[i], id2code))
axarr[1][2].set_title("prediction mask")
for cl in range(len(color_labels)):
if cl > 5:
axarr[1][cl - 5].imshow(y_pred[i, :, :, cl])
axarr[1][cl - 5].set_title("layer " + id2code[cl + 1])
else:
axarr[0][cl].imshow(y_pred[i, :, :, cl])
axarr[0][cl].set_title("layer " + id2code[cl + 1])
plt.show()
# Model learning
result = model.fit(train_inputs,
train_labels,
validation_split=0.2,
batch_size=14,
epochs=300,
callbacks=callbacks)
model.save('models/' + model_name + '.model')
y_pred = model.predict(test_inputs)
y_predi = y_pred
for i in range(TEST_LENGTH):
img_is = (test_inputs[i])
seg = y_predi[i]
fig = plt.figure(figsize=(1, 3))
ax = fig.add_subplot(1, 3, 1)
ax.imshow(img_is)
ax.set_title("original")
ax = fig.add_subplot(1, 3, 2)
ax.imshow(heatmap_to_rgb(y_predi[i], id2code))
ax.set_title("predicted class")
ax = fig.add_subplot(1, 3, 3)
ax.imshow(heatmap_to_rgb(test_labels[i], id2code))
ax.set_title("true class")
plt.show()
def add_masks(self, pred):
return heatmap_to_rgb(pred, id2code)