improvement: remove unnecessary expand and squeeze dimension and intr…

…oduce image send in a batch mode

svnh_semi_supervised_client.py

@@ -18,52 +18,77 @@
 from tensorflow_serving.apis import prediction_service_pb2
 from tensorflow.contrib.util import make_tensor_proto
 
+from tensorflow.contrib.util import make_tensor_proto
+
+from os import listdir
+from os.path import isfile, join
+
 
 def parse_args():
-    parser = ArgumentParser(description="Request a TensorFlow server for a prediction on the image")
-    parser.add_argument("-s", "--server",
-                        dest="server",
+    parser = ArgumentParser(description='Request a TensorFlow server for a prediction on the image')
+    parser.add_argument('-s', '--server',
+                        dest='server',
                         default='172.17.0.2:9000',
-                        help="prediction service host:port")
-    parser.add_argument("-i", "--image",
-                        dest="image",
-                        default="",
-                        help="path to image in JPEG format",)
+                        help='prediction service host:port')
+    parser.add_argument('-i', '--image_path',
+                        dest='image_path',
+                        default='',
+                        help='path to images folder',)
+    parser.add_argument('-b', '--batch_mode',
+                        dest='batch_mode',
+                        default='true',
+                        help='send image as batch or one-by-one')
     args = parser.parse_args()
 
     host, port = args.server.split(':')
 
-    return host, port, args.image
+    return host, port, args.image_path, args.batch_mode == 'true'
 
 
 def main():
     # parse command line arguments
-    host, port, image = parse_args()
+    host, port, image_path, batch_mode = parse_args()
 
     channel = implementations.insecure_channel(host, int(port))
     stub = prediction_service_pb2.beta_create_PredictionService_stub(channel)
+
+    filenames = [(image_path + '/' + f) for f in listdir(image_path) if isfile(join(image_path, f))]
+    files = []
+    imagedata = []
+    for filename in filenames:
+        f = open(filename, 'rb')
+        files.append(f)
 
-    # Send request
-    with open(image, 'rb') as f:
-        # See prediction_service.proto for gRPC request/response details.
         data = f.read()
+        imagedata.append(data)
 
-        start = time.time()
+    start = time.time()
 
+    if batch_mode:
+        print('In batch mode')
         request = predict_pb2.PredictRequest()
-
-        # Call GAN model to make prediction on the image
         request.model_spec.name = 'gan'
         request.model_spec.signature_name = 'predict_images'
-        request.inputs['images'].CopyFrom(make_tensor_proto(data, shape=[1]))
 
-        result = stub.Predict(request, 60.0)  # 60 secs timeout
-
-        end = time.time()
-        time_diff = end - start
+        request.inputs['images'].CopyFrom(make_tensor_proto(imagedata, shape=[len(imagedata)]))
 
+        result = stub.Predict(request, 60.0)
         print(result)
-        print('time elapased: {}'.format(time_diff))
+    else:
+        print('In one-by-one mode')
+        for data in imagedata:
+            request = predict_pb2.PredictRequest()
+            request.model_spec.name = 'gan'
+            request.model_spec.signature_name = 'predict_images'
+
+            request.inputs['images'].CopyFrom(make_tensor_proto(data, shape=[1]))
+
+            result = stub.Predict(request, 60.0)  # 60 secs timeout
+            print(result)
+
+    end = time.time()
+    time_diff = end - start
+    print('time elapased: {}'.format(time_diff))
 
 
 if __name__ == '__main__':

svnh_semi_supervised_model_saved.py

@@ -41,11 +41,6 @@ def preprocess_image(image_buffer):
     # adjust_* ops all require this range for dtype float.
     image = tf.image.convert_image_dtype(image, dtype=tf.float32)
 
-    # Networks accept images in batches.
-    # The first dimension usually represents the batch size.
-    # In our case the batch size is one.
-    image = tf.expand_dims(image, 0)
-
     # Finally, rescale to [-1,1] instead of [0, 1)
     image = tf.subtract(image, 0.5)
     image = tf.multiply(image, 2.0)
@@ -65,8 +60,6 @@ def main(_):
         tf_example = tf.parse_example(serialized_tf_example, feature_configs)
         jpegs = tf_example['image/encoded']
         images = tf.map_fn(preprocess_image, jpegs, dtype=tf.float32)
-        images = tf.squeeze(images, [0])
-        # now the image shape is (1, ?, ?, 3)
 
         # Create GAN model
         z_size = 100