def predict_images(test_data_path=None, video=None): save_dir = os.path.join('saved_model', element) tf.reset_default_graph() with tf.Session() as sess: # Predict the logits input_image, keep_prob, vgg_layer3_out, vgg_layer4_out, vgg_layer7_out = load_vgg( sess, vgg_path) nn_last_layer = layers(vgg_layer3_out, vgg_layer4_out, vgg_layer7_out, num_classes) logits = tf.reshape(nn_last_layer, (-1, num_classes)) saver = tf.train.Saver() last_checkpoint = helper.get_latest_checkpoint_number(save_dir) - 1 if video is None: saver.restore( sess, os.path.join('saved_model', element, 'model-' + str(last_checkpoint))) print("Restored the saved Model in save_model") helper.pred_samples(runs_dir, test_data_path, sess, image_shape, logits, keep_prob, input_image, crops, restrict_prediction) else: save_dir = os.path.join('saved_model', test_data_path) last_checkpoint = helper.get_latest_checkpoint_number(save_dir) - 1 saver.restore( sess, os.path.join('saved_model', test_data_path, 'model-' + str(last_checkpoint))) return helper.get_binary_seg(video, sess, image_shape, logits, keep_prob, input_image, crops)
def opt_predict_images(test_data_path=None, video=None): frozen_graph = os.path.join('optimised_model', element, 'graph.pb') # frozen_graph = os.path.join('frozen_model', element, 'saved_model.pb') tf.reset_default_graph() with tf.gfile.GFile(frozen_graph, "rb") as f: graph_def = tf.GraphDef() graph_def.ParseFromString(f.read()) with tf.Graph().as_default() as graph: tf.import_graph_def(graph_def, input_map=None, return_elements=None, name="") input_image = graph.get_tensor_by_name('image_input:0') keep_prob = graph.get_tensor_by_name('keep_prob:0') logits = graph.get_tensor_by_name('logits:0') sess = tf.Session(graph=graph) # with tf.Session(graph=graph) as sess: if video is None: helper.pred_samples(runs_dir, test_data_path, sess, image_shape, logits, keep_prob, input_image, crops, restrict_prediction) else: return helper.get_binary_seg(video, sess, image_shape, logits, keep_prob, input_image, crops)
def predict_images(test_data_path, print_speed=False): num_classes = 2 image_shape = (160, 576) runs_dir = './outputs' vgg_path = os.path.join('./data', 'vgg') with tf.Session() as sess: input_image, keep_prob, vgg_layer3_out, vgg_layer4_out, vgg_layer7_out = load_vgg( sess, vgg_path) nn_last_layer = layers(vgg_layer3_out, vgg_layer4_out, vgg_layer7_out, num_classes) logits = tf.reshape(nn_last_layer, (-1, num_classes)) saver = tf.train.Saver() saver.restore(sess, model_path) print("Restored the saved Model in file: %s" % model_path) helper.pred_samples(runs_dir, test_data_path, sess, image_shape, logits, keep_prob, input_image, print_speed)
def predict_images(test_data_path, print_speed=False): num_classes = 2 image_shape = (160, 576) runs_dir = '/home/shuijing/Desktop/ece498sm_project/runs' # Path to vgg model vgg_path = os.path.join('/home/shuijing/Desktop/ece498sm_project/', 'vgg') with tf.Session() as sess: # Predict the logits input_image, keep_prob, vgg_layer3_out, vgg_layer4_out, vgg_layer7_out = load_vgg(sess, vgg_path) nn_last_layer = layers(vgg_layer3_out, vgg_layer4_out, vgg_layer7_out, num_classes) logits = tf.reshape(nn_last_layer, (-1, num_classes)) # Restore the saved model saver = tf.train.Saver() saver.restore(sess, model_path) print("Restored the saved Model in file: %s" % model_path) # Predict the samples helper.pred_samples(runs_dir, test_data_path, sess, image_shape, logits, keep_prob, input_image, print_speed)