# io.write_tif(im_path, lb*255, data['geotransform']
# [i], data['geoprojection'][i], data['size'][i])
# # cv2.imwrite(im_path,lb*255)
# # merging = []
# # output_vrt = os.path.join(path_data, 'merged.vrt')
# # for root, dirs, files in os.walk(path_predict):
# # for file in files:
# # if ".tif" in file:
# # merging.append(file)
# # gdal.BuildVRT(output_vrt, merging, options=gdal.BuildVRTOptions(
# # srcNodata=-9999, VRTNodata=-9999))
# # Merging all the tif datasets
# logger.info('Merging tiled dataset')
# io.merge_tile(file_output, predict_image)
# # Converting raster to Vector
# logging.info('Converting Raster to vector')
# output_format = 'shp'
# io.raster2vector(file_output, os.path.dirname(file_output), output_format)
# # Post Processing shp to axis aligned bounding box
# postprocess.aabbox(os.path.dirname(file_output), output_format)
# Saving to accuracy.json
io.tojson(accuracy, os.path.join(path_result, 'accuracy.json'))
logger.info('Completed')
sys.exit()
path_save_callback = os.path.join(config.path_weight,
'weights.{epoch:02d}-{val_loss:.2f}.hdf5')
saving_model = keras.callbacks.ModelCheckpoint(path_save_callback,
monitor='val_loss',
verbose=0,
save_best_only=False,
save_weights_only=True,
mode='auto',
period=5)
# fit the unet with the actual image, train_image
# and the output, train_label
history = unet_model.fit_generator(generator=training_generator,
epochs=config.epoch,
workers=3,
validation_data=validation_generator,
callbacks=[csv_logger, saving_model])
# Saving path of weigths saved
logging.info('Saving model')
unet_model.save(os.path.join(config.path_weight, 'final.hdf5'))
# Getting timings
end_time = time.time() - st_time
timing['Total Time'] = str(end_time)
# Saving to JSON
io.tojson(timing, os.path.join(config.path_model, 'Timing.json'))
logging.info('Completed')
sys.exit()
os.path.join(save_model_lo, 'M_%s_%s.h5' % (str(k), str(j))))
# Saving path of weigths saved
logger.info('Saving weights')
path_weight = os.path.join(save_weight_lo,
'W_%s_%s.h5' % (str(k), str(j)))
umodel.save_weights(path_weight)
# Counting number of loops
count = count + 1
end_loop = time.time()
# Getting timings
timing['loop_%s_%s' % (str(k), str(j))] = end_loop - st_loop
io.tojson(timing, os.path.join(result_lo, 'Timing.json'))
# Clearing memory
train_image = []
train_label = []
end_time = time.time() - st_time
timing['Total Time'] = str(end_time)
# Saving to JSON
io.tojson(timing, os.path.join(result_lo, 'Timing.json'))
# model.evaluate(x=vali_images, y=vali_label, batch_size=32, verbose=1)#, sample_weight=None, steps=None)
# model.predict( vali_images, batch_size=32, verbose=1)#, steps=None)
logger.info('Completed')
sys.exit()
file_skeleton = []
for j in range(len(path_merged)):
temp = join(path_skeleton, basename(path_merged[j]))
file_skeleton.append(temp)
_ = postprocess.skeletonize(path_merged[j], temp)
# Skeletonization completed
timing[current_process[-1]] = mtime.time() - time
# Converting raster to Vector
time = mtime.time()
print('Converting Raster to vector')
path_vector = join(path_merged_prediction, 'vector')
checkdir(path_vector)
file_vector = []
for j in range(len(file_skeleton)):
temp = file_skeleton[j]
path_r2v = io.raster2vector(temp, path_vector, output_format)
# Vectorization completed
timing[current_process[-1]] = mtime.time() - time
# Saving to JSON
io.tojson(timing, join(path_result, 'Timing.json'))
print('Process Completed')
sys.exit()