Sample Applications

Basic demo apps for each mode have been provided with the Object Library. The demos are written in C++ and run on Linux. The demos take no additional command line arguments, and the output of each demo is a streaming video.
Once started, the demos run as long as the ESC key has not been pressed.

SDKRecognitionApp (or_tutorial_1_gui)

This is the single recognition demo app. Its goal is to demonstrate the most basic steps to activate and display recognition results.

  • The app displays the object with the largest probability and the probability itself (and not all of the objects returned by the classifier).
  • This demo perceives the whole frame as its ROI.

SDKLocalizationApp (or_tutorial_2_gui)

This is the localization demo app. Its goal is to demonstrate the most basic steps to activate and display localization results.

  • The localization algorithm is turned on/off by using the left mouse button.
  • The sample displays the bounding box of all the found objects (not ignoring threshold of course) together with the recognized objects inside them, and the probability those objects were recognized with. 5. This demo perceives the whole frame as its ROI.

SDKTrackingApp (or_tutorial_3_gui)

This is the tracking demo app. Its goal is to demonstrate the most basic steps to activate and display tracking results.

  • The tracked ROI is defined by clicking the left mouse button, dragging it, and then clicking the left button again.
  • The sample displays the tracked ROI each frame.

SDKLocalizationAndTrackingApp (or_tutorial_4_gui)

This sample app is designed to provide the developer the best way of performing Object Localization, followed by tracking. The combination of localization and tracking enables the user to know the objects and their location in real-time. The flow of the sample app is as follows:

  • The application starts with a video feed. No Object Library algorithm is active yet.
  • Clicking with the left mouse button on the screen initiates localization on the current frame and tracking on the consecutive frames.
  • It is advised that the camera be held still during localization to avoid tracking failure.
  • Localization will be re-run once all objects that are being tracked are lost or the user again clicks in the screen.

SDKYORTool

  • This is the YOR demo app. Its goal is to demonstrate the most basic steps to activate and display recognition result on newly-trained classes.
  • Meanwhile, the app displays the object with the largest probability and the probability itself (and not all of the objects returned by the classifier).
  • This demo works on the displayed ROI, that is, the object to recognize should fit within the ROI.
For more complete information about compiler optimizations, see our Optimization Notice.