ADLINK Technology EVA SDK User manual

EVA SDK User's Manual
Manual Rev.: 1.2
Revision Date: July 20, 2021
Part Number: 50M-00010-1020

ii Preface
Preface
Copyright
Copyright © 2021 ADLINK Technology, Inc. This document contains proprietary information protected by copyright.
All rights are reserved. No part of this manual may be reproduced by any mechanical, electronic, or other means in
any form without prior written permission of the manufacturer.
Disclaimer
The information in this document is subject to change without prior notice in order to improve reliability, design, and
function and does not represent a commitment on the part of the manufacturer. In no event will the manufacturer
be liable for direct, indirect, special, incidental, or consequential damages arising out of the use or inability to use
the product or documentation, even if advised of the possibility of such damages.
Trademarks
Product names mentioned herein are used for identification purposes only and may be trademarks and/or
registered trademarks of their respective companies.
Revision History
Revision Description Date
1.0 Initial release 2020-09-30
1.1 Release for EVA SDK R3 (Windows) 2021-04-12
1.2 Release for EVA SDK R3.5 2021-07-20

EVA SDK User's Manual
Preface iii
Table of Contents
Preface................................................................................................................................ii
1Introduction................................................................................................................1
2Setting Environment Variables.................................................................................5
3Image and Video Capture Plugins............................................................................7
3.1 Element Descriptions.....................................................................................................7
4Image Processing Plugins......................................................................................33
4.1 Element Descriptions...................................................................................................35
5Image Analytic Plugins ...........................................................................................75
5.1 Element Descriptions...................................................................................................76
6Connector Plugins.................................................................................................103
6.1 Element Descriptions.................................................................................................103
7Convert for OpenVINO Model...............................................................................105
7.1 Model Optimizer Developer Guide.............................................................................105
7.2 Converting a Caffe* Model.........................................................................................105
7.3 Converting a TensorFlow* Model...............................................................................105
7.4 Converting an ONNX* Model.....................................................................................105
7.5 Converting a Tensorflow YoloV3 Model.....................................................................105
8Convert for TensorRT Model................................................................................107
8.1 Build TensorRT Engine Parameters..........................................................................107
8.2 Convert Tensorflow models.......................................................................................108
8.3 Samples.....................................................................................................................108
Safety Instructions.........................................................................................................112
Getting Service...............................................................................................................113

iv Preface
This page intentionally left blank.

EVA SDK User's Manual
Introduction 1
1 Introduction
(EVA Architecture)
The ADLINK Edge Vision Analytics (EVA) SDK provides an integrated development environment (IDE) for
developers wanting to build computer vision and video analytic solutions with deep learning technology. The SDK
provides the required building blocks to help with the implementation of edge inference solutions. Customers can
select either the Intel® Distribution of OpenVINO™ Toolkit, NVIDIA TensorRT™, or both to accelerate deep
learning inference. This SDK can help customers easily migrate or integrate heterogeneous hardware platforms.
ADLINK offers a selection of pre-verified platforms with specific hardware configurations optimized for vision and
video analytic applications requiring little additional engineering effort by developers. The EVA SDK integrates and
verifies different building blocks required for Edge Vision Analytic applications. The building blocks are
implemented as GStreamer plugins. Customers can select a specific plugin based on their requirements, utilizing it
in their development pipeline to quickly build their application. These plugins can be categorized as follows.
•Image/video capture: Supports different image capture solutions with different camera standards.
•Image processing: Provides the required algorithms needed before or after the inference.
•Image analytics: Provides an optimized INTEL/NVIDIA inference engine to accelerate deep learning inference.
•Connector plugin: Provides network connections for uploading images and inference results to a network
service.
All the plugins can be easily assembled to become a pipeline that can form an application. There are three kinds of
plugins: source providers (image/video capture), filters (image processing/image analytics), and data consumers
(connector plugins). Source providers play a leading role in the pipeline, transmitting data to an intermediate filter
that deals with the data for specific processing. The data can then be consumed by the final plugins for data
storage and transmission.
Under the EVA platform, the image capture plugin provides the video stream data to the image preprocessing
plugin, as shown in the figure below. The image processing plugin then delivers the processed data to the AI
inference engine, and the AI inference engine delivers the result. After the intermediate filter process, the data is
published to the data river via a plugin, where other edge devices like OPC UA or robots can subscribe to it.

2 Introduction
The EVA SDK supports application development in C/C++ as well as in Python, with the ability to implement
customized plugins in both programming languages. (Sample code is included in the EVA SDK package.)
The following use scenarios demonstrate how the EVA SDK can be used to easily create vision analytic
applications.
Use Scenario 1: Single pipeline in one application
This scenario demonstrates a simple edge AI application. With the configured pipeline, the image inference can be
made with a single image source on an ADLINK edge AI platform camera (e.g. Neon).
1. Application (1) captures the image from a GigE camera
2. Application (1) does AI inference on Nvidia TensorRT via adrt plugin
3. Application (1) does the inference result interpretation with the adtranslator plugin
Use Scenario 2: Multiple pipelines in one application
This scenario demonstrates an edge AI application. With multiple pipelines, the image inference can be captured
with multiple image sources and the AI inference performed on both Nvidia and Intel hardware with an ADLINK
edge AI platform (e.g. EOS-i).
1. Application (1) captures the image from the IP camera
2. Application (1) decodes the RTSP via NV H.264 HW accelerator

EVA SDK User's Manual
Introduction 3
3. Application (1) does the AI inference for camera 1 on OpenVINO via the advino plugin
4. Application (1) does the inference result interpretation with the adtranslator plugin
5. Application (1) captures the image from the GigE camera
6. Application (1) does the AI inference for the GigE camera on TensorRT via the adrt plugin
7. Application (1) does the inference result interpretation with the adtranslator plugin
Notes:
•The examples using gst-launch in this User’s Manual are for reference only. The “property” values, such as the
file name and path, must be modified according to the actual system environment.
•Support for the elements in the examples are operating system dependent. For example, ‘xvimagesink’ is only
supported on Linux. It is recommended to use the ‘glimagesink’ or ‘d3dvideosink’ elements instead. Note that
only one glimagesink can be used in a pipeline.

4 Introduction
This page intentionally left blank.

EVA SDK User's Manual
Setting Environment Variables 5
2 Setting Environment Variables
Before using EVA SDK, the environment must be installed and set up. For installation details, refer to the EVA
Installation Guide.
On Linux (include x86 and ARM), after installing EVA SDK and required software, open a terminal and run the
following command to set the environment variables.
$ source /opt/adlink/eva/scripts/setup_eva_envs.sh
If you have changed the install path (INSTALL_DIR), be sure to specify the path.
$ source INSTALL_DIR/scripts/setup_eva_envs.sh
The script will set up the environment variables of the following installed software.
•OpenVINO
•MediaSDK
•Pylon
•ADLINK EVA SDK
If the software has no corresponding libraries, the script will not set up the corresponding environment variables.
Note: The environment variables are removed when closing the terminal.
On Windows, after installing EVA SDK and required software, there are two ways to execute the EVA IDE.
Method 1: Manually run the environment variable settings.
Open a Windows command prompt and then run the following commands to set the environment variables; then
enter the GStreamer command or your application to execute.
> C:\ADLINK\gstreamer\setupvars.bat
> C:\ADLINK\eva\scripts\setup_eva_envs.bat
Note: The environment variables are removed when closing the command prompt.
Method 2: Run the desktop shortcut.
On Windows, run EVACMD on the desktop. The shortcut will automatically set the environment and open the
command prompt. You can then enter the GStreamer command.

6 Setting Environment Variables
This page intentionally left blank.

EVA SDK User's Manual
Image and Video Capture Plugins 7
3 Image and Video Capture Plugins
The following table lists the elements of the supported image and video capture plugins.
Element
Name Plugin Name Vendor Linxu/ARM Version Windows Version Description
pylonsrc Pylonsrc PlayGineering 3.5 3.5
A GStreamer element that uses
Basler's USB3 Vision cameras to
capture images
filesrc coreelements GStreamer 1.14.5 1.16.2 Reads from an arbitrary point in a file
rtspsrc rtsp GStreamer 1.14.5 1.16.2
Receives network data via RTSP
(RFC 2326)
v4l2src video4linux2 GStreamer 1.14.5 1.16.2
Reads frames from a Video4Linux2
device
multifilesrc multifile GStreamer 1.14.5 1.16.2
Reads buffers from sequentially
named files
ksvideosrc winks GStreamer - 1.16.2
Provides low-latency video capture
from WDM cameras on Windows
hikplugin hikplugin ADLINK 3.5 3.5
A GStreamer element that uses
Hikrobot's USB & GigE vision
cameras to capture images
flirplugin flirplugin ADLINK 3.5 3.5
A GStreamer element that uses
FLIR's USB & GigE vision cameras to
capture images
3.1 Element Descriptions
This section describes each element of the video capture plugin in greater detail.
3.1.1 pylonsrc
This is a GStreamer element that uses Basler's USB3 Vision cameras and Basler’s GigE Vision cameras for image
capture.
When multiple cameras are available, use the camera parameter to specify the specific camera for image capture.
To find out the IDs for each camera, launch the pipeline without specifying the camera parameter, and the plugin
will output the camera IDs.
The width and height parameters are used to set a custom resolution. To find out the maximum resolution of the
camera, refer to the camera's technical specifications, or launch the plugin with logging tuned to loglevel4
(GST_DEBUG=pylonsrc:4). Note that the camera saves the manually set resolution. Launching a pipeline with a
camera while specifying a resolution will result in the following runs using the same resolution unless the device is
reconnected or a different resolution is specified.
Resolution offsets can be specified to move the image. Possible offsets are calculated by max(width|height)-
current(width|height). The parameters to specify offsets are offsetx and offsety. To center the image,
use centerx and centery. Note that setting the centering parameters will cause the plugin to ignore the offset
values.
The ADLINK EVA SDK uses the pylonsrc source code included in gst-plugins-vision.
https://github.com/joshdoe/gst-plugins-vision/tree/master/sys/pylon

8 Image and Video Capture Plugins
For more details, refer to the gst-pylonsrc documentation at https://gitlab.com/zingmars/gst-pylonsrc/-
/blob/master/README.MD. Note that use ‘pixel-format’ property instead of ‘imageformat’ property when
referencing this document.
Notes:
1. Download and install the ADLINK EVA SDK to access the pylonsrc plugin. For more details, refer to the
ADLINK EVA Installation Guide.
2. The downstream elements may waste a lot of time handling frames. If the camera's frame rate is too fast, the
pipeline will drop a lot of frames, resulting in rendering delays. However, the rendering and AI inference
performance of each platform are not the same. It is recommended to reduce the frame rate of the camera
source or image resolution to make the rendered screen output smoother.
3.1.1.1 Examples Using gst-launch
Displaying a pylon video with ycbcr422_8 format.
gst-launch-1.0 pylonsrc pixel-format=ycbcr422_8 ! videoconvert ! xvimagesink
Recording a pylon video at 150fps.
gst-launch-1.0 pylonsrc limitbandwidth=false sensorreadoutmode=fast fps=150 !
videoconvert ! matroskamux ! filesink location='recording.mkv'
Setting the resolution width and height to 960x540 and displaying the pylon video.
gst-launch-1.0 pylonsrc width= 960 height=540 ! videoconvert ! xvimagesink
Centering the pylon video and display.
gst-launch-1.0 pylonsrc centerx=true centery=true ! videoconvert ! xvimagesink
Note: It is recommended to use xvimagesink to display the window on Linux and use glimagesink or d3dvideosink
to display the window on Windows.
3.1.1.2 Hierarchy
GObject
╰──GInitiallyUnowned
╰──GstObject
╰──GstElement
╰──GstBaseSrc
╰──GstPushSrc
╰──GstPylonsrc

EVA SDK User's Manual
Image and Video Capture Plugins 9
3.1.1.3 Pad Templates
src
ANY
Presence – always
Direction – src
3.1.1.4 Properties
camera
“camera” gint
(Number) Camera ID as defined by Basler's API. If only one camera is connected this parameter will be ignored
and the camera will be used. If there are multiple cameras and this parameter is not defined, the plugin will
output a list of available cameras and their IDs. Note that if there are multiple cameras available to the API and
the camera parameter is not defined then this plugin will not run. Range: 0 to 100
Flags: Read, Write
Default value: 9999
height
“height” gint
(Pixel) The height of the picture. Note that the camera will remember this setting, and will use values from the
previous runs if you relaunch without specifying this parameter. Reconnect the camera or use the reset
parameter to reset. Range: 0 to 10000
Flags: Read, Write
Default value: 0
width
“width” gint
(Pixels) The width of the picture. Note that the camera will remember this setting, and will use values from the
previous runs if you relaunch without specifying this parameter. Reconnect the camera or use the reset
parameter to reset. Range: 0 to 10000
Flags: Read, Write
Default value: 0
binningh
“binningh” gint
(Pixels) The number of pixels to be binned in the horizontal direction. Note that the camera will remember this
setting, and will use values from the previous runs if you relaunch without specifying this parameter. Reconnect
the camera or use the reset parameter to reset. Range: 1 to 6
Flags: Read, Write
Default value: 1

10 Image and Video Capture Plugins
binningv
“binningv” gint
(Pixels) The number of pixels to be binned in the vertical direction. Note that the camera will remember this
setting, and will use values from the previous runs if you relaunch without specifying this parameter. Reconnect
the camera or use the reset parameter to reset. Range: 1 to 6
Flags: Read, Write
Default value: 1
limitbandwidth
“limitbandwidth” gboolean
(true/false) Bandwidth limit mode. CAUTION! Disabling this will allow the camera to reach higher frames per
second, but can damage the camera. Running the plugin without specifying this parameter will reset the value
stored on the camera to `true`.
Flags: Read, Write
Default value: true
maxbandwidth
“maxbandwidth” gint64
(Bytes per second) This property sets the maximum bandwidth the camera can use. The camera will only use
as much as it needs for the specified resolution and framerate. This setting will have no effect if limitbandwidth is
set to off. Note that the camera will remember this setting, and will use values from the previous runs if you
relaunch without specifying this parameter. Reconnect the camera or use the reset parameter to reset. Range:
0 to 999999999
Flags: Read, Write
Default value: 0
sensorreadoutmode
“sensorreadoutmode” gchararray
(normal/fast) This property changes the sensor readout mode. Fast will allow for faster framerates but might
cause quality loss. It might be required that either increasing the maximum bandwidth or disabling bandwidth
limits be configured for this to cause any noticeable change. Running the plugin without specifying this
parameter will reset the value stored on the camera to "normal".
Flags: Read, Write
Default value: "normal"
acquisitionframerateenable
“acquisitionframerateenable” gboolean
(true/false) Enables the use of custom fps values. This parameter will be set to true if the fps property is set.
Running the plugin without specifying this parameter will reset the value stored on the camera to false.
Flags: Read, Write
Default value: false

EVA SDK User's Manual
Image and Video Capture Plugins 11
fps
“fps” gdouble
(Frames per second) Sets the framerate of the video coming from the camera. Setting the value too high might
cause the plugin to crash. Note that if the pipeline causes the computer to hang, or stall, then the resulting video
will not be in the resolution that was set. Setting this parameter will set acquisitionframerateenable to true. The
value of this parameter will be saved to the camera, but it will have no effect unless either this or the
acquisitionframerateenable parameters are set. Reconnect the camera or use the reset parameter to reset.
Range: 0.0 to 1024.0
Flags: Read, Write
Default value: 0.0
lightsource
“lightsource” gchararray
(off, 2800k, 5000k, 6500k) Changes the color balance settings to those defined by the presets. For best results,
select a color balance setting closest to the environment's lighting. Running the plugin without specifying this
parameter will reset the value stored on the camera to "5000k".
Flags: Read, Write
Default value: "5000k"
autoexposure
“autoexposure” gchararray
(off, once, continuous) Controls whether the camera will try to adjust the exposure settings. Setting this
parameter to anything but "off" will override the exposure parameter. Running the plugin without specifying this
parameter will reset the value stored on the camera to "off".
Flags: Read, Write
Default value: "off"
exposure
“exposure” gdouble
(Microseconds) Exposure time for the camera in microseconds, but only has an effect if autoexposure is set to
off (default). Higher numbers will cause a lower frame rate. Note that the camera will remember this setting, and
will use values from the previous runs if you relaunch without specifying this parameter. Reconnect the camera
or use the reset parameter to reset. Range: 0.0 to 1000000.0
Flags: Read, Write
Default value: 0.0
autowhitebalance
“autowhitebalance” gchararray
(off, once, continuous) Controls whether the camera will try to adjust the white balance settings. Setting this
parameter to anything but "off" will override the exposure parameter. Running the plugin without specifying this
parameter will reset the value stored on the camera to "off".
Flags: Read, Write
Default value: "off"

12 Image and Video Capture Plugins
balancered
“balancered” gdouble
Specifies the red color balance. The autowhitebalance setting must be set to "off" for this property to have any
effect. Note that this value gets saved on the camera, and running this plugin again without specifying this value
will cause the previous value to be used. Use the reset parameter or reconnect the camera to reset. Range: 0.0
to 15.9
Flags: Read, Write
Default value: 999
balancegreen
“balancegreen” gdouble
Specifies the green color balance. The autowhitebalance setting must be set to "off" for this property to have
any effect. Note that this value gets saved on the camera, and running this plugin again without specifying this
value will cause the previous value to be used. Use the reset parameter or reconnect the camera to reset.
Range: 0.0 to 15.9
Flags: Read, Write
Default value: 999
balanceblue
“balanceblue” gdouble
Specifies the blue color balance. The autowhitebalance setting must be set to "off" for this property to have any
effect. Note that this value gets saved on the camera, and running this plugin again without specifying this value
will cause the previous value to be used. Use the reset parameter or reconnect the camera to reset. Range: 0.0
to 15.9
Flags: Read, Write
Default value: 999
colorredhue
“colorredhue” gdouble
Specifies the red hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 999
colorredsaturation
“colorredsaturation” gdouble
Specifies the red saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 0.0

EVA SDK User's Manual
Image and Video Capture Plugins 13
coloryellowhue
“coloryellowhue” gdouble
Specifies the yellow hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 999
coloryellowsaturation
“coloryellowsaturation” gdouble
Specifies the yellow saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 999
colorgreenhue
“colorgreenhue” gdouble
Specifies the green hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 999
colorgreensaturation
“colorgreensaturation” gdouble
Specifies the green saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 999
colorcyanhue
“colorcyanhue” gdouble
Specifies the cyan hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 999

14 Image and Video Capture Plugins
colorcyansaturation
“colorcyansaturation” gdouble
Specifies the cyan saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 999
colorbluehue
“colorbluehue” gdouble
Specifies the blue hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 0.0
colorbluesaturation
“colorbluesaturation” gdouble
Specifies the blue saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 999
colormagentahue
“colormagentahue” gdouble
Specifies the magenta hue. Note that this value gets saved on the camera, and running this plugin again without
specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the camera
to reset. Range: -4.0 to 3.9
Flags: Read, Write
Default value: 999
colormagentasaturation
“colormagentasaturation” gdouble
Specifies the magenta saturation. Note that this value gets saved on the camera, and running this plugin again
without specifying this value will cause the previous value to be used. Use the reset parameter or reconnect the
camera to reset. Range: 0.0 to 1.9
Flags: Read, Write
Default value: 999

EVA SDK User's Manual
Image and Video Capture Plugins 15
autogain
“autogain” gchararray
(off, once, continuous) Controls whether the camera will try to adjust the gain settings. Setting this parameter to
anything but "off" will override the exposure parameter. Running the plugin without specifying this parameter will
reset the value stored on the camera to "off".
Flags: Read, Write
Default value: "off"
gain
“gain” gdouble
(dB) Sets the gain added on the camera before sending the frame to the computer. The value of this parameter
will be saved to the camera, but it will be set to 0 every time this plugin is launched without specifying gain, or it
will be overridden if the autogain parameter is set to anything that is not "off". Reconnect the camera or use the
reset parameter to reset the stored value. Range: 0.0 to 12.0
Flags: Read, Write
Default value: 0.0
blacklevel
“blacklevel” gdouble
(DN) Sets the stream's black level. This parameter is processed on the camera before the picture is sent to the
computer. The value of this parameter will be saved to the camera, but it will be set to 0 every time this plugin is
launched without specifying this parameter. Reconnect the camera or use the reset parameter to reset the
stored value. Range: 0.0 to 63.75
Flags: Read, Write
Default value: 0.0
gamma
“gamma” gdouble
Sets the gamma correction value. This parameter is processed on the camera before the picture is sent to the
computer. The value of this parameter will be saved to the camera, but it will be set to 1.0 every time this plugin
is launched without specifying this parameter. Reconnect the camera or use the reset parameter to reset the
stored value. Range: 0.0 to 3.9
Flags: Read, Write
Default value: 1.0
reset
“reset” gchararray
(off, before, after). Controls whether or when the camera's settings will be reset. Setting this to "before" will wipe
the settings before the camera initialization begins. Setting this to "after" will reset the device once the pipeline
closes. This can be useful for debugging or when using the camera with other software that does not reset the
camera settings before use (such as PylonViewerApp).
Flags: Read, Write
Default value: "off"

16 Image and Video Capture Plugins
testimage
“testimage” gint
(1 to 6) Specifies a test image to show instead of a video stream. It is useful for debugging and is disabled by
default. Range: 0 to 6
Flags: Read, Write
Default value: 0
continuous
“continuous” gboolean
(true/false) Used to switch between triggered and continuous mode. Set to “false” to switch to triggered mode.
Flags: Read, Write
Default value: True
pixel-format
“pixel-format” gchararray
(Mono8 / BayerBG8 / BayerGR8 / BayerRG8 / BayerGB8 / RGB8 / BGR8 / YCbCr422_8 / YUV422Packed).
Determines the pixel format for sending frames. The default is auto which will query the camera for supported
pixel formats and allow GStreamer to negotiate the format with downstream elements.
Flags: Read, Write
Default value: "auto"
userid
“userid” gchararray
(<string>) Sets the device custom ID.
Flags: Read, Write
Default value: NULL
demosaicing
“demosaicing” gboolean
(true/false) Switches between simple and Basler's demosaicing (PGI) mode. This does not work if bayer output
is used.
Flags: Read, Write
Default value: False
noisereduction
“noisereduction” gdouble
Specifies the amount of noise reduction to apply. To use this, Basler's demosaicing mode must be enabled.
Setting this will enable demosaicing mode. Range: 0.0 to 2.0
Flags: Read, Write