ADLINK Technology NEON-2000-JT2-X Series, User Manual

Page: 34 / 53

24 Getting Started
Leading
EDGE COMPUTING
3.5.2 Inference Concepts and Sample Code
This section illustrates the main concepts of inference with sample
files and also provides a link to examples from NVIDIA. The
default path of the sample files is
/home/Desktop/Capture_and_
Inference_Sample .
The following folders and files in this directory are essential for
image capture and inference tasks:

System requirements and sample build-up procedures:
/Capture_and_Inference_Sample /README.md
 Pre-training model for image classification:
/Capture_and_Inference_Sample/data/networks

Sample code:
/Capture_and_Inference_Sample/examples/imagenet-cam-
era/imagenet-camera.cpp
 Capturing images:
/Capture_and_Inference_Sample/utils/camera/indCam-
era.cpp
 Execution file:
/Capture_and_Inference_Sample/build/aarch64/bin/ima-
genet-camera
To learn more about image sensors, samples and documents are
included in
/opt/pylon5 . For more samples relating to image infer-
ence, NVIDIA provides examples at https://github.com/dusty-nv/
jetson-inference.
3.5.3 Image Inference Procedure
Use these steps to set up a streaming camera and to display and
classify an image inference.
1. Create a camera.
ADLINK provides indcamera.cpp and indcamera.h , which pack-
age several pylon APIs for developers to create custom applica-
tions for the NEON-2000-JT2-X.
2. Create an ImageNet.