Пример #1
0
 def test_inference(self):
     gqcnn = NeuralNetWork(self.config)
     config_ = self.config['gqcnn_config']
     with gqcnn.graph.as_default():
         image = tf.placeholder(tf.float32, [
             None, config_['im_width'], config_['im_height'],
             config_['im_channels']
         ],
                                name='image_input')
         pose = tf.placeholder(tf.float32, [None, config_['pose_dim']],
                               name='pose_input')
     out = gqcnn.inference(image, pose)
     self.assertIsNotNone(out)
Пример #2
0
 def test_dataset(self):
     cfg = self.config['training']
     network = NeuralNetWork(self.config, training=True)
     train = GQCNNTraing(self.config, network, DATA_PATH, OUT_PATH)
     with tf.Session(graph=train._network.graph) as sess:
         sess.run(train._train_iterator.initializer)
         im, pose, label = sess.run(train._train_iterator.get_next())
         im_shape = (cfg['train_batch_size'], cfg['im_height'],
                     cfg['im_width'], cfg['im_channels'])
         pose_shape = (cfg['train_batch_size'], cfg['pose_len'])
         label_shape = (cfg['train_batch_size'], )
         self.assertEqual(im.shape, im_shape)
         self.assertEqual(pose.shape, pose_shape)
         self.assertEqual(label.shape, label_shape)
Пример #3
0
 def test_load_npz(self):
     gqcnn = NeuralNetWork(self.config)
     with tf.Session(graph=gqcnn.graph, config=self.gpu_config) as sess:
         gqcnn.load_weights(sess, SAVE_PATH)
Пример #4
0
 def test_save(self):
     gqcnn = NeuralNetWork(self.config)
     with tf.Session(graph=gqcnn.graph, config=self.gpu_config) as sess:
         gqcnn.load_weights(sess, MODEL_PATH, remap=True)
         gqcnn.save_to_npz(sess, SAVE_PATH)
Пример #5
0
 def test_load_ckpt(self):
     gqcnn = NeuralNetWork(self.config)
     with tf.Session(graph=gqcnn.graph, config=self.gpu_config) as sess:
         gqcnn.load_weights(sess, MODEL_PATH, remap=True)
Пример #6
0
 def test_create(self):
     gqcnn = NeuralNetWork(self.config)
     self.assertIsNotNone(gqcnn)
Пример #7
0
def main():
    config_logging(TEST_LOG_FILE)
    config = load_config(TEST_CFG_FILE)
    network = NeuralNetWork(config, training=True)
    train = GQCNNTraing(config, network, DATA_PATH, OUT_PATH)
    train.optimize(50)
Пример #8
0
 def test_create(self):
     network = NeuralNetWork(self.config, training=True)
     train = GQCNNTraing(self.config, network, DATA_PATH, OUT_PATH)
     # train.optimize(10)
     self.assertIsNotNone(train)