Modify OpenVINO and Streaming Python doc (#1414)

* Add System requirements into OpenVINO doc and OpenVINO example.
* Modify Streaming Python doc.
* Fix path problem in autograd example.

docs/docs/APIGuide/PipelineAPI/inference.md

@@ -14,11 +14,19 @@ Inference Model is a package in Analytics Zoo aiming to provide high-level APIs
 
 **OpenVINO requirements:**
 
+[System requirements](https://software.intel.com/en-us/openvino-toolkit/documentation/system-requirements):
+
+    Ubuntu 16.04.3 LTS (64 bit)
+    Windows 10 (64 bit)
+    CentOS 7.4 (64 bit)
+    macOS 10.13, 10.14 (64 bit)
+
+Python requirements:
+
     tensorflow>=1.2.0
     networkx>=1.11
     numpy>=1.12.0
     protobuf==3.6.1
-    onnx>=1.1.2
 
 **Supported models:**
 

docs/docs/ProgrammingGuide/inference.md

@@ -28,11 +28,19 @@ Inference Model is a package in Analytics Zoo aiming to provide high-level APIs
 
 **OpenVINO requirements:**
 
+[System requirements](https://software.intel.com/en-us/openvino-toolkit/documentation/system-requirements):
+
+    Ubuntu 16.04.3 LTS (64 bit)
+    Windows 10 (64 bit)
+    CentOS 7.4 (64 bit)
+    macOS 10.13, 10.14 (64 bit)
+
+Python requirements:
+
     tensorflow>=1.2.0
     networkx>=1.11
     numpy>=1.12.0
     protobuf==3.6.1
-    onnx>=1.1.2
 
 **Java**
 

pyzoo/zoo/examples/autograd/README.md

@@ -8,8 +8,8 @@ Follow the instructions [here](https://analytics-zoo.github.io/master/#PythonUse
 You can easily use the following commands to run this example:
 ```
 export SPARK_DRIVER_MEMORY=2g
-python custom.py
-python customloss.py
+python path/to/custom.py
+python path/to/customloss.py
 ```
 
 See [here](https://analytics-zoo.github.io/master/#PythonUserGuide/run/#run-after-pip-install) for more running guidance after pip install.
@@ -24,12 +24,12 @@ ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master ${MASTER}\
     --driver-memory 2g \
     --executor-memory 2g \
-    custom.py 
+    path/to/custom.py
 ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master ${MASTER}\
     --driver-memory 2g \
     --executor-memory 2g \
-    customloss.py 
+    path/to/customloss.py
 ```
 See [here](https://analytics-zoo.github.io/master/#PythonUserGuide/run/#run-without-pip-install) for more running guidance without pip install.
 

pyzoo/zoo/examples/openvino/README.md

@@ -7,6 +7,19 @@ to make inferences with OpenVINO toolkit as backend using Analytics Zoo, which d
 ## Install or download Analytics Zoo
 Follow the instructions [here](https://analytics-zoo.github.io/master/#PythonUserGuide/install/) to install analytics-zoo via __pip__ or __download the prebuilt package__.
 
+[OpenVINO System requirements](https://software.intel.com/en-us/openvino-toolkit/documentation/system-requirements):
+
+    Ubuntu 16.04.3 LTS (64 bit)
+    Windows 10 (64 bit)
+    CentOS 7.4 (64 bit)
+    macOS 10.13, 10.14 (64 bit)
+
+OpenVINO Python requirements:
+
+    tensorflow>=1.2.0
+    networkx>=1.11
+    numpy>=1.12.0
+    protobuf==3.6.1
 
 ## Model and Data Preparation
 1. Prepare a pre-trained TensorFlow object detection model. You can download from [tensorflow detection model zoo](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md).
@@ -21,7 +34,7 @@ In this example, we use `frozen_inference_graph.pb` of the `faster_rcnn_resnet10
 ```bash
 export SPARK_DRIVER_MEMORY=10g
 image_path=directory path that contain images
-model_path=path to frozen_inference_graph.pb
+model_path=dir contains frozen_inference_graph.pb and pipeline.config
 
 python predict.py --image ${image_path} --model ${model_path}
 ```
@@ -35,7 +48,7 @@ export SPARK_HOME=the root directory of Spark
 export ANALYTICS_ZOO_HOME=the directory where you extract the downloaded Analytics Zoo zip package
 MASTER=local[*]
 image_path=directory path that contain images
-model_path=path to frozen_inference_graph.pb
+model_path=dir contains frozen_inference_graph.pb and pipeline.config
 
 ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master $MASTER \

pyzoo/zoo/examples/openvino/predict.py

@@ -16,6 +16,7 @@
 
 import sys
 import numpy as np
+from os.path import join
 from optparse import OptionParser
 
 from zoo.common.nncontext import init_nncontext
@@ -41,7 +42,9 @@
                            resize_height=600, resize_width=600).get_image().collect()
     input_data = np.concatenate([image.reshape((1, 1) + image.shape) for image in images], axis=0)
     model = InferenceModel()
-    model.load_tf(options.model_path, backend="openvino", model_type=options.model_type)
+    model.load_tf(join(options.model_path, "frozen_inference_graph.pb"),
+                  backend="openvino", model_type=options.model_type,
+                  ov_pipeline_config_path=join(options.model_path, "pipeline.config"))
     predictions = model.predict(input_data)
     # Print the detection result of the first image.
     print(predictions[0])

pyzoo/zoo/examples/streaming/objectdetection/README.md

@@ -1,7 +1,7 @@
 # Analytics Zoo Streaming Object Detection
 Imagining we have pre-trained model and image files in file system, and we want to detect objects in these images. In streaming case, it's not an easy task to read image files with help of a third part framework (such as HDFS or Kafka). To simplify this example, we package image paths into text files. Then, these image paths will be passed to executors through streaming API. Executors will read image content from file systems, and make prediction. The predicted results (images with boxes) will be stored to output dir.
 
-So, there are two applications in this example: image_path_writer and streaming_object_detection. ImagePathWriter will package image paths into text files. Meanwhile, StreamingObjectDetection read image path from those text files, then read image content and make prediction.
+So, there are two applications in this example: image_path_writer and streaming_object_detection. image_path_writer will package image paths into text files. Meanwhile, streaming_object_detection read image path from those text files, then read image content and make prediction.
 
 ## Environment
 * Python (2.7, 3.5 or 3.6)
@@ -19,26 +19,26 @@ Make sure all nodes can access image files, model and text files. Local file sys
 ```
 MASTER=...
 model=... // model path. Local file system/HDFS/Amazon S3 are supported
-streamingPath=... // text files location. Only local file system is supported
-output=... // output path of prediction result. Only local file system is supported
+streaming_path=... // text files location. Only local file system is supported
+output_path=... // output path of prediction result. Only local file system is supported
 ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master ${MASTER} \
     --driver-memory 5g \
     --executor-memory 5g \
     streaming_object_detection.py \
-    --streamingPath ${streamingPath} --model ${model} --output ${output}
+    --streaming_path ${streaming_path} --model ${model} --output_path ${output_path}
 ```
 
 2. Start image_path_writer
 ```
 MASTER=...
-imageSourcePath=... // image path. Only local file system is supported
-streamingPath=... // text files. Only local file system is supported
+img_path=... // image path. Only local file system is supported
+streaming_path=... // text files. Only local file system is supported
 ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master ${MASTER} \
     --driver-memory 5g \
     image_path_writer.py \
-    --streamingPath ${streamingPath} --imageSourcePath ${imageSourcePath}
+    --streaming_path ${streaming_path} --img_path ${img_path}
 ```
 
 ## Results

pyzoo/zoo/examples/streaming/textclassification/README.md

@@ -21,14 +21,14 @@ nc -lk [port]
 ```
 MASTER=...
 model=... // model path. Local file system/HDFS/Amazon S3 are supported
-indexPath=... // word index path. Local file system/HDFS/Amazon S3 are supported
+index_path=... // word index path. Local file system/HDFS/Amazon S3 are supported
 port=... // The same port with nc command
 ${ANALYTICS_ZOO_HOME}/bin/spark-submit-with-zoo.sh \
     --master ${MASTER} \
     --driver-memory 2g \
     --executor-memory 5g \
     streaming_text_classification.py \
-    --model ${model} --indexPath ${indexPath} --port ${port}
+    --model ${model} --index_path ${index_path} --port ${port}
 ```
 
 3. TERMINAL 1: Input text in Netcat