fix readme (#1862)

* unify cmd line, fix readme
* handle reading errors

Author: vladimir-dudnik

demos/python_demos/image_translation_demo/README.md

@@ -25,18 +25,18 @@ python3 cocosnet_demo.py -h
 The command yields the following usage message:
 
 ```
-usage: image_translation_demo.py [-h] -t TRANSLATION_MODEL
-                                 [-s SEGMENTATION_MODEL] [-ii INPUT_IMAGES]
+usage: image_translation_demo.py [-h] -m_trn TRANSLATION_MODEL
+                                 [-m_seg SEGMENTATION_MODEL] [-ii INPUT_IMAGES]
                                  [-is INPUT_SEMANTICS] -ri REFERENCE_IMAGES
                                  [-rs REFERENCE_SEMANTICS] -o OUTPUT_DIR
                                  [-d DEVICE]
 
 Options:
   -h, --help            Show this help message and exit.
-  -t TRANSLATION_MODEL, --translation_model TRANSLATION_MODEL
+  -m_trn TRANSLATION_MODEL, --translation_model TRANSLATION_MODEL
                         Required. Path to an .xml file with a trained
                         translation model
-  -s SEGMENTATION_MODEL, --segmentation_model SEGMENTATION_MODEL
+  -m_seg SEGMENTATION_MODEL, --segmentation_model SEGMENTATION_MODEL
                         Optional. Path to an .xml file with a trained
                         semantic segmentation model
   -ii INPUT_IMAGES, --input_images INPUT_IMAGES
@@ -69,35 +69,35 @@ To run the demo, you can use public or pre-trained models. You can download the
 
 There are two ways to use this demo:
 
-1. To use only translation model.
-   You can use the following command run demo on CPU using CoCosNet as translation model:
+1. Run with segmentation model in addition to translation model. You should use only models trained on ADE20k dataset.     Example: [hrnet-v2-c1-segmentation](../../../models/public/hrnet-v2-c1-segmentation/hrnet-v2-c1-segmentation.md).
+   In this case only input and reference images are required without any masks.
+   Segmentation masks will be generated via segmentation model.
+
+   You can use the following command to run demo on CPU using cocosnet and hrnet-v2-c1-segmentation models:
 
    ```
-   python3 cocosnet_demo.py \
+   python3 image_translation_demo.py \
        -d CPU \
-       -t  <path_to_translation_model>/CoCosNet.xml \
-       -is <path_to_semantic_mask_of_image>/input_mask.png \
-       -ri <path_to_exemplar_image>/reference_image.jpg \
-       -rs <path_to_exemplar_semantic>/reference_mask.png
+       -m_trn <path_to_translation_model>/cocosnet.xml \
+       -m_seg <path_to_segmentation_model>/hrnet-v2-c1-segmentation.xml \
+       -ii <path_to_input_image>/input_image.jpg \
+       -ri <path_to_exemplar_image>/reference_image.jpg
    ```
 
-   > **NOTE**: For segmentation masks you should use mask (with shape: [height x width]) that specifies class for each pixel. Number of classes is 151 (from ADE20k), where '0' - background class.
-
-2. To use the segmentation model in addition to translation. You should use only models trained on ADE20k dataset.     Example: [hrnet-v2-c1-segmentation](../../../models/public/hrnet-v2-c1-segmentation/hrnet-v2-c1-segmentation.md).
-   In this case user have to set input image and reference image without any masks.
-   Segmentation masks will be generated via segmentation model.
-
-   You can use the following command run demo on CPU using CoCosNet as translation model:
+2. Run with only translation model.
+   You can use the following command to run demo on CPU using cocosnet as translation model:
 
    ```
-   python3 cocosnet_demo.py \
+   python3 image_translation_demo.py \
        -d CPU \
-       -t <path_to_translation_model>/CoCosNet.xml \
-       -s <path_to_segmentation_model>/Seg.xml \
-       -ii <path_to_input_image>/input_image.jpg \
-       -ri <path_to_exemplar_image>/reference_image.jpg
+       -m_trn <path_to_translation_model>/cocosnet.xml \
+       -is <path_to_semantic_mask_of_image>/input_mask.png \
+       -ri <path_to_exemplar_image>/reference_image.jpg \
+       -rs <path_to_exemplar_semantic>/reference_mask.png
    ```
 
+   > **NOTE**: For segmentation masks you should use mask (with shape: [height x width]) that specifies class for each pixel. Number of classes is 151 (from ADE20k), where '0' - background class.
+
 ## Demo Output
 
 The results of the demo processing are saved to a folder that is specified by the parameter `output_dir`.

demos/python_demos/image_translation_demo/image_translation_demo.py

@@ -30,10 +30,10 @@ def build_argparser():
     args = parser.add_argument_group('Options')
     args.add_argument('-h', '--help', action='help', default=SUPPRESS,
                       help='Show this help message and exit.')
-    args.add_argument("-t", "--translation_model",
+    args.add_argument("-m_trn", "--translation_model",
                       help="Required. Path to an .xml file with a trained translation model",
                       required=True, type=str)
-    args.add_argument("-s", "--segmentation_model",
+    args.add_argument("-m_seg", "--segmentation_model",
                       help="Optional. Path to an .xml file with a trained semantic segmentation model",
                       type=str)
     args.add_argument("-ii", "--input_images",
@@ -104,13 +104,29 @@ def main():
         samples = [number_of_objects * [''], input_semantics, reference_images, reference_semantics]
     for input_img, input_sem, ref_img, ref_sem in zip(*samples):
         if use_seg:
-            input_sem = get_mask_from_image(cv2.imread(input_img), seg_model)
-            ref_sem = get_mask_from_image(cv2.imread(ref_img), seg_model)
+            in_img = cv2.imread(input_img)
+            if in_img is None:
+                raise IOError('Image {} cannot be read'.format(input_img))
+            input_sem = get_mask_from_image(in_img, seg_model)
+            r_img = cv2.imread(ref_img)
+            if r_img is None:
+                raise IOError('Image {} cannot be read'.format(ref_img))
+            ref_sem = get_mask_from_image(r_img, seg_model)
         else:
-            input_sem = cv2.imread(input_sem, cv2.IMREAD_GRAYSCALE)
+            input_sem_file = input_sem
+            input_sem = cv2.imread(input_sem_file, cv2.IMREAD_GRAYSCALE)
+            if input_sem is None:
+                raise IOError('Image {} cannot be read'.format(input_sem_file))
+            ref_sem_file = ref_sem
             ref_sem = cv2.imread(ref_sem, cv2.IMREAD_GRAYSCALE)
+            if ref_sem is None:
+                raise IOError('Image {} cannot be read'.format(ref_sem_file))
         input_sem = preprocess_semantics(input_sem, input_size=gan_model.input_semantic_size)
-        ref_img = preprocess_image(cv2.imread(ref_img), input_size=gan_model.input_image_size)
+        ref_img_file = ref_img
+        ref_img = cv2.imread(ref_img_file)
+        if ref_img is None:
+            raise IOError('Image {} cannot be read'.format(ref_img_file))
+        ref_img = preprocess_image(ref_img, input_size=gan_model.input_image_size)
         ref_sem = preprocess_semantics(ref_sem, input_size=gan_model.input_semantic_size)
         input_dict = {
             'input_semantics': input_sem,