Applies to: FLEX AI, P-AI Series Cameras

Summary:

This article describes how to move through a training video and mark objects. The goal is to take a video clip and pause at several points in the video and annotate (draw boxes around) the objects you are teaching FLEX AI to detect.

Step By Step Guide:

Annotating Objects

To annotate objects:

1. Pause the video on any frame that contains the object of interest.
   
   
   
2. Draw a detection box around every instance of the object.
   There can be multiple instances of an object and each instance should have its own detection box. 
   
   
   
3. Click the Tighten button (or the keyboard shortcut T) to resize the detection boxes.
   This means you can save time by roughly drawing the boxes and then tightening.
   

   Step 1: Roughly draw a box around the object.
   Step 2: Click the Tighten button (or press T) to quickly get you close.
   There will be times that FLEX AI does not properly capture the true edges of the object.    Step 3: Drag the edges/corners of a detection box to finely edit.

4. To remove a detection box, select the appropriate box and click the Delete key on your keyboard.
   
   
5. Click Save (or the keyboard shortcut S) to save the marked detection boxes.
   
   
6. Once you have drawn the necessary detection boxes, click the Train button.
   Your project will be unavailable during the training period which typically takes 15 to 30 minutes depending on the number of annotations.
   

NOTE: Since training is done in the cloud, you can work on any number of other projects 
while a project is training.

NOTE: The blue detection boxes also appear in the This Frame’s Detections area after you 
save. It is recommended that a minimum of 50 detection boxes be defined before your 
initial object training is complete. An AI model can include detection boxes from 
multiple video clips.

Keyboard Shortcuts

You can save time and effort when navigating video, marking, tightening, and saving annotations by using the following keyboard shortcuts.