Please enable JavaScript to view this site.

IDS peak 2.10.0 / uEye+ firmware 3.34

Navigation: E: AI-assisted image analysis with IDS peak

Executing CNNs in IDS peak Cockpit

Scroll Previous Top Next More

IDS Peak comfortSDK, genericSDK, IPL, and AFL developer manuals are external documents. Please contact us if you need them.

You can load and run neural network (CNN) in IDS peak Cockpit. You can execute networks for classification as well as for object detection. After image acquisition by the camera, the images are transferred to the AI library in host processing and analyzed. The entire image is used for the analysis. The result is displayed in the camera image in the IDS peak Cockpit.

As the images are processed in the host processing, it is possible to execute both classification and object detection on the same image.

Prerequisite: The CNNs were trained for use with IDS peak. For this purpose, you can use the AI Vision Studio IDS lighthouse.

Nets for classification have the file extension *.pcla

Nets for object detection have the file extension *.pdet

hint_info

Notes for using AI-assisted image analysis

The inference is executed on the host system and therefore the performance depends on the host system.

The CPU load can be very high depending on the CNN used and the host system.

If you use multiple networks at the same time, the CPU load will be correspondingly higher.

The inference is processed synchronously via the IDS peak comfortSDK interface, i.e. the function only returns after an image has been analyzed.

Optionally, after installing IDS peak, you find three simple neural networks for testing under "<IDS peak installation directory>/cnn". These networks were created for test purposes and are not intended for productive use.

CNN-name

Type

Short desctiption

cats_dogs_classification.pcla

Classification

Classifies dogs or cats

item_classification.pcla

Classification

Classifies simple items like pens, car keys, wallets, cameras etc.

person_detection.pdet

Object detection

Detects people

© 2024 IDS Imaging Development Systems GmbH