In our macOS CCTV software SecuritySpy, deciding whether to trigger recording and actions (e.g. notifications) is a two-step process. Firstly, a motion-detection algorithm determines whether there is some interesting movement happening. Next, the image is cropped to the area of movement and passed to a neural-network classifier to determine whether there is an object of interest present, like a human, vehicle or animal.

The classifier in SecuritySpy is based on a ResNet-50 neural network, custom-trained on over 100,000 real CCTV images from thousands of different cameras. It is a highly optimised, powerful network, producing very accurate results (around 97% for humans and vehicles).

In contrast, until now, the motion detection algorithm has been fairly simple. It has worked via pixel-based comparisons between successive video frames, with some additional filtering to help ignore extraneous movement caused by noise, transient motion and lighting changes.

This standard motion detection is fast and very effective in controlled environments that don’t suffer from a lot of extraneous motion. However, it is less effective in more challenging outdoor environments, where movement caused by shadows, swaying plants, snow, rain and so on can cause a high number of false-positive detections. In such situations, it maintains high sensitivity (ability to detect true motion when it occurs), but suffers from poor specificity (ability to accurately ignore extraneous movement).

SecuritySpy’s classifier is able to filter out virtually all these false-positives, however even if only 3% of false-positive detections get through the classifier, in particularly challenging situations where the motion detector is generating large numbers of false-positives, inevitably there will still be a significant number getting through and causing triggers.

Standard motion detection pipeline: with False Positives (FP) greatly outnumbering True Positives (TP), even though the classifier is very accurate, a significant number still get through and cause triggers

This is solved by SecuritySpy’s new neural-network based motion detector. Built from the ground up and trained on tens of thousands of image sequences, this is a bespoke lightweight CNN (Convolutional Neural Network) featuring residual connections and multi-scale pooling for real-time motion detection and location. It outputs highly accurate predictions, even in challenging outdoor situations.

The new neural network is specifically trained to ignore common extraneous motion. In particular, rain and snow at night gets brightly lit by cameras’ built-in infrared lighting, causing large white constantly-moving streaks across the image. Other motion like trees swaying, lighting shifts, and camera shaking are also effectively ignored.

With the output of the new AI-powered motion detector being far more accurate, far fewer false-positives are sent to the classifier, which in turn will then cause far fewer false-positive triggers of recording or actions.

AI motion detection pipeline: with far fewer false-positives getting through to the classifier stage, virtually all false-positive triggers can be avoided

Performance Considerations

Due to the powerful Neural Engine module built into Apple Silicon chips, the new AI motion detector runs just as fast, if not faster, than the standard algorithm (in around 0.7 ms per frame). Therefore, on Apple Silicon Macs, SecuritySpy will default to this method for all cameras.

However, the ability to execute neural networks on Intel Macs varies significantly. SecuritySpy will attempt to first benchmark the performance of the new AI, and will use this information to automatically decide how many cameras to apply it to.

You can check how fast each neural network runs on your Mac by selecting the menu item at SecuritySpy > Debug > Run AI Test. Here’s an example:

AI Test window in SecuritySpy. Bonus feature: drag and drop an image file onto any of the image squares to test the classifier on your own images!

If this indicates that all your cameras can be covered by the new AI motion detection algorithm, then you don’t need to take any action; all your cameras will automatically benefit from the new AI.

However, if this indicates that the number of cameras the AI algorithm can be applied to is lower than the number of cameras that you are using, then only a subset of your cameras will benefit from the new AI. By default, SecuritySpy applies this preferentially to the first cameras in the list.

Alternatively, you can specify yourself which cameras should use Standard vs. AI motion detection – in this way, you can apply the AI only to the most important cameras for you, or only those that are currently producing significant numbers of false-positive detections. You will find this option under Settings > Cameras > Triggers > Advanced Options:

AI-powered motion detection was introduced in SecuritySpy version 6.10, and is available on all Macs that are running macOS 10.15 or later.