Applies to Models: P-AI Series Cameras

Summary:

This article covers improving the AI Object Detection model of your uploaded videos. This process will result in more accurate detections, fewer false positives, and fewer false negatives.

Step By Step Guide:

To improve the AI Object Detection model:

1. Click the Improve button (towards the top right of the video player) after you have reviewed the latest model.
   
   
   
2. Select a video to start.
   FLEX AI will suggest the top 10 frames crucial for highlighting your object(s) to enhance your object detection model. Markers represent each frame the application is asking you to revisit.
   
   
   
3. Draw a detection box around every instance of the object.
   Remember to save your detection boxes.
   
   
4. Click the Object Not Present button (towards the bottom right of the video player) if the object of interest is not in the frame.
   
   
   
5. Use the arrows or the orange dots to advance to different/suggested frames.
   
   
   
6. Complete all 10 suggested frames and repeat for all videos to be included in the next training.
   
   
7. Click the Add More button to manually look through the video and add detection boxes, if the suggested frames are not enough (ex: different angles and perspectives are needed in the training set).
   
   
8. Click Train when you are ready.
   The current limit for a (1) year license is to have a maximum of 100 total trainings or improvements.

NOTE: All 10 suggested frames must be completed per video for them to be included 
in the training. Detection boxes on any video where all 10 frames are not marked 
will be discarded and not included.