def main(): # capture the config path from the run arguments # then process the json configration file # try: data_loader = DataLoader(data_dir, config) data_loader.load_directory('.tif') data_loader.create_np_arrays() data_loader.create_data_label_pairs() preptt = PrepTrainTest(config, data_loader) for data_label_pair in data_loader.data_label_pairs: x_data = data_label_pair[0][data_loader] y_true = data_label_pair[1][data_loader.data_label_pairs[i][1][:, :, 0]] preptt.add_data(x_data, y_true) # Create the experiments dirs create_dirs([config.summary_dir, config.checkpoint_dir, config.input_dir]) # Create tensorflow session sess = tf.Session() # Create instance of the model you want model = PopModel(config) # Load model if exist model.load(sess) # Create Tensorboard logger logger = Logger(sess, config) logger.log_config() # Create your data generator data = DataGenerator(config, preptraintest = preptt) data.create_traintest_data() # Create trainer and path all previous components to it trainer = PopTrainer(sess, model, data, config, logger) # Train model trainer.train()
def main(): # capture the config path from the run arguments # then process the json configration file # try: args = get_args() if args.config != 'None': config = process_config(args.config) else: config = process_config(os.path.join(config_dir, 'example.json')) data_loader = DataLoader(data_dir) data_loader.load_directory('.tif') data_loader.create_np_arrays() preptt = PrepTrainTest(data_loader.arrays[0], data_loader.arrays[1], config.batch_size, config.chunk_height, config.chunk_width) prepd = PrepData(data_loader.arrays[0], data_loader.arrays[1], config.batch_size, config.chunk_height, config.chunk_width) # create the experiments dirs create_dirs([config.summary_dir, config.checkpoint_dir]) # create tensorflow session sess = tf.Session() # create instance of the model you want model = PopModel(config) #load model if exist model.load(sess) # create your data generator data = DataGenerator(config, preptt, prepd) data.create_traintest_data() data.create_data() # Create Tensorboard logger logger = Logger(sess, config) # Create trainer and path all previous components to it tester = PopTrainer(sess, model, data, config, logger) # Test model tester.test()
def main(): data_loader = DataLoader(data_dir, config) data_loader.load_directory('.tif') data_loader.create_np_arrays() prepd = PrepData(config, data_loader) start_raster = data_loader.arrays[-1] prepd.add_data(start_raster) # Create the experiments output dir create_dirs([ config.output_dir, config.output_pred_dir, config.output_dif_dir, config.output_eval_dir ]) # Create tensorflow session sess = tf.Session() # Create instance of the model model = PopModel(config) # Load model if exist model.load(sess) # Create data generator data = DataGenerator(config, prepdata=prepd) start_raster = data.prepdata.x_data[0][:, :, 0] prev_raster = start_raster with sess: rasters = [] for k in range(config.num_outputs): data.prepdata.output_nr = k data.create_data() cur_row = 0 cur_col = 0 chunk_height = data.prepdata.chunk_height chunk_width = data.prepdata.chunk_width chunk_rows = data.prepdata.chunk_rows chunk_cols = data.prepdata.chunk_cols output_raster = np.empty( (chunk_rows * chunk_height, chunk_cols * chunk_width)) # Predicting for each batch for i in range(data.batch_num): #y_pred = sess.run(model.y, feed_dict={model.x: data.input[0][i]}) y_pred, y_pred_chunk = sess.run( [model.y, model.y_chunk], feed_dict={ model.x: data.input[0][i], model.x_pop_chunk: data.x_chunk_pop[0][i], model.x_proj: config.pop_proj[k], model.x_cur_pop: data.prepdata.x_cur_pop[0] }) y_pred = y_pred.reshape(config.batch_size, chunk_height, chunk_width) for j in range(config.batch_size): if chunk_cols == cur_col: # Change to new row and reset column if it reaches the end cur_row += 1 cur_col = 0 output_raster[cur_row * chunk_height: (cur_row + 1) * chunk_height, cur_col * chunk_width: (cur_col + 1) * chunk_width] = \ y_pred[j, :, :] cur_col += 1 # Removes null-cells at the start of the array output_raster = output_raster[data.prepdata.offset_rows:, data.prepdata.offset_cols:] # Makes sure the right amount of null-cells are removed from the end of the array if data.prepdata.row_null_cells == 0 and data.prepdata.col_null_cells == 0: pass elif data.prepdata.row_null_cells == 0: output_raster = output_raster[:, :-data.prepdata. col_null_cells] elif data.prepdata.col_null_cells == 0: output_raster = output_raster[:-data.prepdata. row_null_cells, :] else: output_raster = output_raster[:-data.prepdata.row_null_cells, : -data.prepdata.col_null_cells] # Removes the previous input data and adds the output raster data.prepdata.x_data = [] # Replaces the old population with the predicted one, keeps the other features constant new_input = data_loader.arrays[-1] new_input[:, :, 0] = output_raster[:new_input.shape[0], :new_input. shape[1]] data.prepdata.add_data(new_input) rasters.append(output_raster) print('Min value pop: {}'.format(np.amin(output_raster))) print('Max value pop: {}'.format(np.amax(output_raster))) print('Sum value pop: {}'.format(np.sum(output_raster))) # Calculating back to population # norm_sum = np.sum(output_raster) # final_pop = np.sum(pop_arr_14) # # output_raster = (output_raster / norm_sum) * final_pop print(np.max(output_raster)) print(np.min(output_raster)) print(output_raster.shape) data_writer = DataWriter(data_loader.geotif[0], start_raster, prev_raster, output_raster, config) data_writer.write_outputs() prev_raster = output_raster