# EVA SDK

Edge Vision Analytics (EVA) Software Development Kit (SDK) for ADLINK AI Vision Products

![EVM\nEDGE VISION ANALYTICS\n2 Days\nfor Demo\n2 Weeks\nfor PoC\n2 Months\nto Deploy\nIntroduction](.eva-sdk-datasheet-20210927/50b7bfad6e3e63c1e5c9c36d1580cb182a2017f37ea4af39bf58bbb57957525b.jpg)

# Intr

EVA SDK is a unified edge vision analytics service-ready software platform that enables ADLINK AI hardware, making it easier for users to develop optimized edge AI vision applications by simplifying integration and focusing on essential functionality. Users can leverage readyto-use open-source plugins to facilitate each stage the AI vision project lifecycle, including image capture and processing, AI inference, postprocessing, and analytics. This "One API" framework allows users to successfully build a proof-of-concept in two weeks and speed up mass deployment time.

# Key Features

# The Best PoC Tool for Edge AI Vision Applications

● AI Pipeline Studio
![C:\Users\000\nVideo\nAudio\nAudio Comparison\nAudio Comparison](.eva-sdk-datasheet-20210927/ef82131a4f9f2a1a9d412fc48988f4362d84d5f389a1bbe187a98754a53798bd.jpg)

GUI interface for AI applications, editing, executing, debugging, and performance profiling

● Abundant Ready-to-use Plugins
![The flowchart depicts a system with the following components and connections:\n\n1.  **Input:** On the far left, there is a **white security camera**.\n2.  **Branching:** An arrow extends from the camera and splits into two paths:\n    *   The top path points to a **blue circle** containing the Python snake logo, labeled '**python**'.\n    *   The bottom path points to a **blue hexagon** containing the C++ logo.\n3.  **Convergence and Output:** Arrows extend from both the 'python' circle and the 'C++' hexagon to the right. These lines merge into a single arrow pointing towards an **illuminated lightbulb** on the far right.](.eva-sdk-datasheet-20210927/85864ae86a84ff33d969a49de80f112fb943ea33609b9c5fd0b6a8ca2d7820bd.jpg)

Hardware codec, image processing, and edge communication, including stream out

● Optimized AI Models
![Blue icon of a microchip with 'AI' text, no additional symbols or text present](.eva-sdk-datasheet-20210927/de506863414789c308d64f73b9525ab995ab6c8eeffceb91145a3b689235f215.jpg)

Object classification, detection, segmentation, and human posture detection

# Easy Migration from Lab Test to Field PoC

● Supports Hybrid AI Inference Engines Simultaneously in a Pipeline
![The image displays the NVIDIA logo set against a solid lime-green background. A white square is positioned in the upper right quadrant, featuring a stylized white spiral design that resembles an eye or swirl. Directly below this square is the text 'nVIDIA' in white sans-serif font, accompanied by a small 'TM' trademark symbol to its right.](.eva-sdk-datasheet-20210927/5db7821a4ab22c45d55f761c5da9f1dae9aa1cb01add2ef36df0f03107b781e3.jpg)

• NVIDIA: GPGPU, Jetson
• Intel: CPU, iGPU, VPU

intel

● Support Multiple Industrial Camera Types
![The image displays a logo featuring the text 'USB' in large, bold letters. The 'U' and 'S' are black, while the 'B' is red, accompanied by a small registered trademark symbol (®) to its right. Below this main text, the word 'VISION' appears in a smaller, black font with a slightly distorted or glitch-like style.](.eva-sdk-datasheet-20210927/b6ec2beb3306d72d33a69e538438884593009f0879d348852c6719a65c8f3379.jpg)

![The image displays the logo for GigE Vision. The text 'GiGE' is shown in the upper portion; the 'Gi' is in black, while the 'GE' is in red, with the 'G' stylized as three horizontal bars. Below this, the word 'VISION' appears in grey capital letters. A small 'TM' symbol is visible to the right of the red 'E'.](.eva-sdk-datasheet-20210927/2892900318dfecb459702dd53a5cfae48360e0377173b4859d53abc0c6cf912e.jpg)

![The image displays the logo for the text 'mipi' in black, lowercase sans-serif letters. Above the text, specifically over the 'pi,' is a semi-circular arc made of small, multicolored dots.](.eva-sdk-datasheet-20210927/9dbfbbf445d7f3f641f36cb2083ab3debcfade98739ebf064cdfcb6f097f57da.jpg)

• Basler, FLIR, Hikrobot GigE and USB3 vision camera
• RTSP, V4L2

● Supports Operation between Different Operating Systems
![The image displays two logos stacked vertically. The top logo features the word 'ubuntu' in black lowercase sans-serif font next to an orange circle icon with a white center. The bottom logo is a blue rectangle containing the Windows icon (four white panes) on the left and the word 'Windows' in white text next to it.](.eva-sdk-datasheet-20210927/624c37754ff767aa9d0b3cfdd2b734f6617605f106f9c8ae6fa98604a9050752.jpg)
No porting effort for platform migration

# Intuitive GUI for Fast and Easy AI Inference Pipeline Development

![Based on the provided image, here is the accurate and concise description of the flowchart/block diagram labeled '3. Edit Window ) Pipeline Workflow Composition'.\n\n**Labeled Blocks:**\n*   `filsr`\n*   `decodebin`\n*   `videoconvert` (top row)\n*   `adrino`\n*   `adtrans_yolo`\n*   `adMetadataViewer`\n*   `videoconvert` (bottom row)\n*   `dsdvideosink`\n\n**Connections:**\n*   `filsr` (output `src`) connects to `decodebin` (input `sink`).\n*   `decodebin` (output `src`) connects to `videoconvert` (input `sink`).\n*   `videoconvert` (output `src`) connects to `adrino` (input `sink`).\n*   `adrino` (output `src`) connects to `adtrans_yolo` (input `sink`).\n*   `adtrans_yolo` (output `src`) connects to `adMetadataViewer` (input `sink`).\n*   `adMetadataViewer` (output `src`) connects to `videoconvert` (input `sink`).\n*   `videoconvert` (output `src`) connects to `dsdvideosink` (input `sink`).\n*   `adtrans_yolo` (output `src`) also connects directly to `dsdvideosink` (input `sink`).](.eva-sdk-datasheet-20210927/a8ee04a7f18152b1c6cab8fb1f69e338422fb965e48addf9a36a312968a8b4fc.jpg)

2. Debug Window Run-time Information Inspection

# Key Functions

![The image displays a black-and-white graphic logo. It features a thick black outline of an eye. Inside the eye, on the left side, is a large, sans-serif capital letter 'G'. To the right of the 'G' are three small triangles stacked vertically, all pointing to the right.](.eva-sdk-datasheet-20210927/5ab33621b3a92d4f0d14e0ba630962f82ff2bdefa1eadac4ba88a68135d3a46b.jpg)
Image Capturing

• Basler USB3/GigE camera
• FLIR USB3/GigE camera
• Hikrobot USB3/GigE camera
• RTSPS stream
• V4L2
• Video/Image files

![The image shows a gray icon depicting two adjacent triangular peaks, resembling mountains, with the right peak being taller than the left. The icon is centered within a square border composed of small gray dashes.](.eva-sdk-datasheet-20210927/f3015cb62b6f12e958b3e21b3fe68e913e4c3323f364391c2f6c2d6428a53011.jpg)
Image Pre-Processing

• NVIDIA H.264/H.256 HW CODEC
• Intel H.264/H.265 HW CODEC
• Pixel format transfer
• Image rotate/flip/crop
• Gamma correction
• Median filter
• Brightness, contrast, hue, saturation

![The image displays a dark gray icon of a microchip or processor centered on a white background. Inside the chip, the letters 'AI' are printed in bold, white, sans-serif font. The chip features rectangular pins protruding from its top, bottom, left, and right sides.](.eva-sdk-datasheet-20210927/adbd631bb366ee19c1831e0e5b69c7948d7cf2f00e474846731d24391b3cbb25.jpg)
AI Inference

• Object Classification
• Object Detection
• Segmentation
• Human pose
• Customized model

![The image displays a black-and-white schematic line drawing. It features four circles arranged in a network-like structure. Three circles are aligned horizontally across the top, and one circle is centered below them. A horizontal wavy line runs across the middle of the image. The three top circles connect to the upper edge of this wavy line via short vertical lines, while the bottom circle connects to the lower edge of the wavy line via a short vertical line. The drawing resembles a bus topology in computer networking.](.eva-sdk-datasheet-20210927/0522161c17ca100429a0c4307d597d84387283219e23e56b6754013be92eb10c.jpg)
Connectivity

• RTSP stream out
• AWS connector
• DDS data river (2021.Q3)
• ROS2 (2021.Q3)
• MS Azure Connector (2021.Q3)

AI Models Verified on EVA

<table><tr><td rowspan="2">Application</td><td rowspan="2">Network</td><td rowspan="2">Original Framework</td><td>TensorRT</td><td>OpenVINO</td></tr><tr><td>Verified</td><td>Verified</td></tr><tr><td rowspan="2">Classification (OK/NG, or sorting)</td><td>GoogLeNet</td><td>Caffe</td><td>V</td><td>V</td></tr><tr><td>MobilenetV2</td><td>ONNX</td><td>V</td><td>V</td></tr><tr><td rowspan="5">Defect/Object Detection</td><td>SSD Inception-V2</td><td>Tensorflow</td><td>V</td><td>-</td></tr><tr><td>Face detection (SSD based)</td><td>Caffe</td><td>X</td><td>V</td></tr><tr><td>Yolov3</td><td>Darknet</td><td>V</td><td>V</td></tr><tr><td>Yolov4</td><td>Darknet</td><td>V</td><td>-</td></tr><tr><td>Yolov3-tiny</td><td>Darknet</td><td>V</td><td>-</td></tr><tr><td>Pose Detection</td><td>OpenPose</td><td>PyTorch</td><td>V</td><td>V</td></tr><tr><td rowspan="2">Segmentation</td><td>FCN-Alexnet</td><td>Caffe</td><td>V</td><td>X</td></tr><tr><td>Intel Segmentation</td><td>Caffe</td><td>X</td><td>V</td></tr></table>

Note: The specific configuration of neural network and original framework is validated and supported by EVA. For other configuration options, please contact ADLINK for support.

Supported AI Vision Platforms & Cameras

<table><tr><td>Product Type</td><td>Model</td></tr><tr><td>AI Smart Camera</td><td>NEON-2000-JNX</td></tr><tr><td rowspan="2">Vision System</td><td>EOS-i6000-P</td></tr><tr><td>EOS-JNX-I / EOS-JNX-G</td></tr></table>
[🔗 Link to the original document](.eva-sdk-datasheet-20210927/eva-sdk-datasheet-20210927.pdf)