In last week’s ‘sneak peak’ at the new features in Kinesense version 3.2, we covered the new options for Multiple Video Exports, Undo Annotations and the FFMpeg update. This week we’ll be looking at the new algorithm options for processing video and how to view them.
Deep learning is a very useful new tool in artificial intelligence. It has now been integrated into Kinesense as an option in the the ‘standard’ algorithm.
When selecting the analysis algorithm select Standard and then go the Advanced Settings. There you will have the option to select Use Deep Learning. Check that box and then you can select the model you want to use. Version 3.2 ships with a model that identifies two categories (Cars and Pedestrians) and we’ll be adding more categories in future.
After you’ve imported your video using the deep learning model you’ll be able to search for the behaviours using filters. The filter wizard will now have an extra page where you can select what type of object you are interested in.
Doorway analysis is another new algorithm that can be selected when importing a video. This analysis aims to detect movement going in and out of doorways and exclude irrelevant movement such as cars passing in front of the door.
You can now also apply a doorway filter to your video after you’ve processed it
Version 3.2 will add a new way to view the results of your import. When importing using face detection, you’ll now be able to view the faces in a grid and sort them by different attributes.
You’ll then be able to select the results you are interested and export them to either the report or tags tabs where you can make your reports.