print "Learning step: " + str(learning_step_current_nr) + "..."
# Get image from training set.
img_train_name = img_train_names[it]
# Load the image and labels of image with name 'img_train_name'
img_train = np.load(TRAINING_AND_TESTING_DATA_ROOT + "img_npy/" + img_train_name)
labels_train = np.load(
TRAINING_AND_TESTING_DATA_ROOT + "labels_npy/" + img_train_name.replace("_img.npy", "_labels.npy")
)
# Preprocess the data before loading it into FCN.
img_train_FCN_ready = lib.rgb_to_bgr(img_train)
img_train_FCN_ready = lib.subtract_bgr_channel_means(
img=img_train_FCN_ready,
blue_mean=channel_means["blue_mean"],
green_mean=channel_means["green_mean"],
red_mean=channel_means["red_mean"],
)
img_train_FCN_ready = img_train_FCN_ready.transpose(2, 0, 1).reshape(1, 3, 500, 500)
labels_train_FCN_ready = labels_train.reshape(1, 1, 500, 500)
# Load image and labels into FCN.
solver.net.blobs["data"].data[...] = img_train_FCN_ready
solver.net.blobs["label"].data[...] = labels_train_FCN_ready
# Make one training step.
solver.step(1)
# Measure and temporarily store the loss every 'loss_interval' training steps.
if it % loss_interval == 0:
loss_temp = float(solver.net.blobs["loss"].data)
# Print current learning step.
print "Learning step: " + str(learning_step_current_nr) + '...'
# Get image from training set.
img_train_name = img_train_names[it]
# Load the image and labels of image with name 'img_train_name'
img_train = np.load(TRAINING_AND_TESTING_DATA_ROOT + 'img_npy/' +
img_train_name)
labels_train = np.load(TRAINING_AND_TESTING_DATA_ROOT + 'labels_npy/' +
img_train_name.replace('_img.npy', '_labels.npy'))
# Preprocess the data before loading it into FCN.
img_train_FCN_ready = lib.rgb_to_bgr(img_train)
img_train_FCN_ready = lib.subtract_bgr_channel_means(img=img_train_FCN_ready,
blue_mean=channel_means['blue_mean'],
green_mean=channel_means['green_mean'],
red_mean=channel_means['red_mean'])
img_train_FCN_ready = img_train_FCN_ready.transpose(2, 0, 1).reshape(1, 3, 500, 500)
labels_train_FCN_ready = labels_train.reshape(1, 1, 500, 500)
# Load image and labels into FCN.
solver.net.blobs['data'].data[...] = img_train_FCN_ready
solver.net.blobs['label'].data[...] = labels_train_FCN_ready
# Make one training step.
solver.step(1)
# Measure and temporarily store the loss every 'loss_interval' training steps.
if it % loss_interval == 0:
loss_temp = float(solver.net.blobs['loss'].data)
loss_current.append(loss_temp)
labels_predicted = np.argmax(scores, 2).astype(np.uint8)
lib.plot_sat_image_and_labels(img=img_cropped,
lab=labels_predicted,
alpha=0.5,
plot_img=True,
store_img=False,
plot_with='pyplot',
dir_name=RESULTS_ROOT,
file_name='test_image_x_' + area_name + RUN_NAME + str(number_of_training_steps),
cmap=cmap1)
img = lib.rgb_to_bgr(img)
blue_mean = channel_means['blue_mean']
green_mean = channel_means['green_mean']
red_mean = channel_means['red_mean']
img = lib.subtract_bgr_channel_means(img, blue_mean, green_mean, red_mean)
tiles_of_image = lib.image_to_tiles(img=img,
hgh=500,
wdt=500)
net.predict([tiles_of_image[0, 0, :, :, :]], oversample=False)
caffe_out = net.blobs['score'].data[0, :, :, :]
asdf = np.argmax(caffe_out, 0).astype(np.uint8)
lib.plot_sat_image_and_labels(img=img_cropped,
lab=asdf,
alpha=0.5,
plot_img=False,
store_img=True,
plot_with='Image',
dir_name=RESULTS_ROOT,