OVEP documentation changes for python and C# samples (#268)

Author: saurabhkale17

c_sharp/OpenVINO_EP/yolov3_object_detection/README.md

@@ -9,7 +9,7 @@ The source code for this sample is available [here](https://github.com/microsoft
 # How to build
 
 ## Prerequisites
-1. Install [.NET 5.0](https://dotnet.microsoft.com/download/dotnet/5.0) or higher and download nuget for your OS (Mac, Windows or Linux). Refer [here](https://onnxruntime.ai/docs/build/inferencing.html#prerequisites-1).
+1. Install [.NET 7.0](https://dotnet.microsoft.com/en-us/download/dotnet/7.0) or higher and download nuget for your OS (Mac, Windows or Linux). Refer [here](https://onnxruntime.ai/docs/build/inferencing.html#prerequisites-1).
 2. [The Intel<sup>®</sup> Distribution of OpenVINO toolkit](https://docs.openvinotoolkit.org/latest/index.html)
 3. Use any sample Image as input to the sample.
 4. Download the latest YOLOv3 model from the ONNX Model Zoo.

python/OpenVINO_EP/tiny_yolo_v2_object_detection/README.md

@@ -9,14 +9,13 @@ The source code for this sample is available [here](https://github.com/microsoft
 # How to build
 
 ## Prerequisites
-1. For Windows, [download OpenVINO package](https://storage.openvinotoolkit.org/repositories/openvino/packages) select appropriate OpenVINO version, windows os and extract it. Run setupvars.bat file which is in the root directory of the extracted OpenVINO package to add OpenVINO libraries to PATH.
-2. Download the latest tinyYOLOv2 model from the ONNX Model Zoo.
+1. Download the latest tinyYOLOv2 model from the ONNX Model Zoo.
    This model was adapted from [ONNX Model Zoo](https://github.com/onnx/models).Download the latest version of the [tinyYOLOv2](https://github.com/onnx/models/tree/master/vision/object_detection_segmentation/tiny-yolov2) model from here.
 
 ## Install ONNX Runtime for OpenVINO™ Execution Provider
-Please install the onnxruntime-openvino python package from [here](https://pypi.org/project/onnxruntime-openvino/1.11.0/). The package for Linux contains prebuilt OpenVINO Libs with ABI 0.
+Please install the onnxruntime-openvino python package from [here](https://pypi.org/project/onnxruntime-openvino). The package for Linux contains prebuilt OpenVINO Libs with ABI 0.
 ```
-pip3 install onnxruntime-openvino
+pip3 install onnxruntime-openvino openvino
 ```
 
 ## Optional Build steps for ONNX Runtime
@@ -48,14 +47,6 @@ python3 tiny_yolov2_obj_detection_sample.py --h
 ```bash
 python3 tiny_yolov2_obj_detection_sample.py --video face-demographics-walking-and-pause.mp4 --model tinyyolov2.onnx --device CPU_FP32
 ```
-Note: Make sure the sample is having below lines to use OpenVINO™ in Windows
-```
-import platform
-
-if platform.system() == "Windows":
-	from openvino import utils
-	utils.add_openvino_libs_to_path()
-```
 
 ## To stop the sample from running
 ```bash

python/OpenVINO_EP/tiny_yolo_v2_object_detection/requirements.txt

@@ -1,9 +1,8 @@
 certifi==2022.5.18.1
 flatbuffers==2.0
-numpy==1.22.4
-onnx==1.11.0
+onnx
 opencv-python==4.5.5.64
 Pillow==9.3.0
-protobuf==4.21.6
+protobuf==3.19.0
 scipy==1.7.3
 typing-extensions==4.2.0

python/OpenVINO_EP/tiny_yolo_v2_object_detection/tiny_yolov2_obj_detection_sample.py

@@ -12,7 +12,7 @@
 import platform
 
 if platform.system() == "Windows":
-    from openvino import utils
+    import onnxruntime.tools.add_openvino_win_libs as utils
     utils.add_openvino_libs_to_path()
 
 # color look up table for different classes for object detection sample
@@ -29,7 +29,7 @@
 
 def parse_arguments():
   parser = argparse.ArgumentParser(description='Object Detection using YOLOv2 in OPENCV using OpenVINO Execution Provider for ONNXRuntime')
-  parser.add_argument('--device', default='CPU_FP32', help="Device to perform inference on 'cpu (MLAS)' or on devices supported by OpenVINO-EP [CPU_FP32, GPU_FP32, GPU_FP16, MYRIAD_FP16, VAD-M_FP16].")
+  parser.add_argument('--device', default='CPU_FP32', help="Device to perform inference on 'cpu (MLAS)' or on devices supported by OpenVINO-EP [CPU_FP32, CPU_FP32, GPU_FP32, GPU_FP16].")
   parser.add_argument('--video', help='Path to video file.')
   parser.add_argument('--model', help='Path to model.')
   args = parser.parse_args()
@@ -150,16 +150,16 @@ def main():
     print("Device type selected is 'cpu' which is the default CPU Execution Provider (MLAS)")
     #Specify the path to the ONNX model on your machine and register the CPU EP
     sess = rt.InferenceSession(args.model, so, providers=['CPUExecutionProvider'])
-  elif (args.device == 'CPU_FP32' or args.device == 'GPU_FP32' or args.device == 'GPU_FP16' or args.device == 'MYRIAD_FP16' or args.device == 'VADM_FP16'):
+  elif (args.device == 'CPU_FP32', args.device == 'CPU_FP16' or args.device == 'GPU_FP32' or args.device == 'GPU_FP16'):
     #Specify the path to the ONNX model on your machine and register the OpenVINO EP
     sess = rt.InferenceSession(args.model, so, providers=['OpenVINOExecutionProvider'], provider_options=[{'device_type' : args.device}])
     print("Device type selected is: " + args.device + " using the OpenVINO Execution Provider")
     '''
     other 'device_type' options are: (Any hardware target can be assigned if you have the access to it)
-    'CPU_FP32', 'GPU_FP32', 'GPU_FP16', 'MYRIAD_FP16', 'VAD-M_FP16'
+    'CPU_FP32', 'CPU_FP16', 'GPU_FP32', 'GPU_FP16'
     '''
   else:
-    raise Exception("Device type selected is not [cpu, CPU_FP32, GPU_FP32, GPU_FP16, MYRIAD_FP16, VADM_FP16]")
+    raise Exception("Device type selected is not [cpu, CPU_FP32, GPU_FP32, GPU_FP16]")
 
   # Get the input name of the model
   input_name = sess.get_inputs()[0].name

python/OpenVINO_EP/yolov4_object_detection/README.md

@@ -16,13 +16,12 @@ The source code for this sample is available [here](https://github.com/microsoft
 # How to build
 
 ## Prerequisites
-1. For Windows, [download OpenVINO package](https://storage.openvinotoolkit.org/repositories/openvino/packages) select appropriate version, windows os and extract it. Run setupvars.bat file which is in the root directory of the extracted OpenVINO package to add OpenVINO libraries to PATH.
-2. Download the latest version of the [YOLOv4](https://github.com/onnx/models/tree/master/vision/object_detection_segmentation/yolov4) model from here.
+1. Download the latest version of the [YOLOv4](https://github.com/onnx/models/tree/master/vision/object_detection_segmentation/yolov4) model from here.
 
 ## Install ONNX Runtime for OpenVINO™ Execution Provider
-Please install the onnxruntime-openvino python package from [here](https://pypi.org/project/onnxruntime-openvino/1.11.0/). The package for Linux contains prebuilt OpenVINO Libs with ABI 0.
+Please install the onnxruntime-openvino python package from [here](https://pypi.org/project/onnxruntime-openvino). The package for Linux contains prebuilt OpenVINO Libs with ABI 0.
 ```
-pip3 install onnxruntime-openvino
+pip3 install onnxruntime-openvino openvino
 ```
 
 ## Optional Build steps for ONNX Runtime
@@ -59,14 +58,6 @@ python3 yolov4.py --device CPU_FP32 --video classroom.mp4 --model yolov4.onnx
 ```
 Note:
 * You can pick different device options to run on OpenVINO™ Execution Provider like GPU_FP32, GPU_FP16 and MYRIAD_FP16.
-* Make sure the sample is having below lines to use OpenVINO™ in Windows
-	```
-	import platform
-
-	if platform.system() == "Windows":
-		from openvino import utils
-		utils.add_openvino_libs_to_path()
-	```
 
 ### Run the sample on default CPU Execution Provider (MLAS)
 ```bash

python/OpenVINO_EP/yolov4_object_detection/requirements.txt

@@ -1,9 +1,8 @@
 certifi==2022.5.18.1
 flatbuffers==2.0
-numpy==1.22.4
-onnx==1.11.0
+onnx
 opencv-python==4.5.5.64
 Pillow==9.3.0
-protobuf==4.21.6
+protobuf==3.19.0
 scipy==1.7.3
 typing-extensions==4.2.0

python/OpenVINO_EP/yolov4_object_detection/yolov4.py

@@ -21,7 +21,7 @@
 import platform
 
 if platform.system() == "Windows":
-    from openvino import utils
+    import onnxruntime.tools.add_openvino_win_libs as utils
     utils.add_openvino_libs_to_path()
 
 def image_preprocess(image, target_size, gt_boxes=None):