I thought I would share something I came across and tested recently. I personally still feel Blue Iris is a better option due to its pricing and customizability, but for those of you with a Synology NAS who like to tinker, this may work if you just need a couple of cameras. I did not come up with this and the original link is below, but added a few notes along the way. Note, this only works on intel-based Synology NAS boxes with a minimum of 4GB of RAM.
Synology Surveillance Station - AI Motion Detection
Synology Surveillance Station - AI Motion Detection
- Setup Action Rules in Surveillance Station for each camera as described in the blog post.
- Install Docker on the NAS.
- Install the Deepquestai/Deepstack (Tag: noavx or cpu-x3-beta) image and obtain free API as described in the blog post. Don’t forget to add the API key after setting up the Docker image by using the link to the NAS as described in the blog post.
- Create the following folders on the NAS: /docker/sssAi/Config and /docker/sssAi/CaptureDir
- Create the cameras.json and settings.json files using code in the blog and upload to the /docker/sssAi/Config directory created in the previous step. A couple notes, tweak the URL so “Incoming” is \”Incoming\” . May also need to switch https:// to http:// and 5001 to 5000 depending on the NAS setup. The settings.json file may include other items for object detection found at this link: Object Detection
- Now with everything in place, go back to Docker and install sssAi as described in the blog post. An additional note, also mount /docker/sssAi/CaptureDir to /CaptureDir
- Under each camera in Surveillance Station, go to Recording Settings --> Schedule and change to either Continuous or Delete entirely. Ensure Motion Detection (Green) is not present on the schedule; otherwise, it will record for both the general motion event from Surveillance Station and the AI recording which is duplicative.
- Under Notifications in Surveillance Station, change the Rules to only receive an alert for Action Rule (System) and not Motion detected (Camera). Tailor this however you like.
- This still depends on either the Camera or Surveillance Station detecting the original motion so tweak the sensitivities and thresholds.
- Because this depends on a motion event from Surveillance Station, if it detects motion such as wind moving tree branches, it will send a snapshot to the AI for analysis. This could prove problematic if you have a camera pointed at the driveway for instance as the AI will detect a car and trigger a recording as it looks at the entire frame. Again, messing with the motion detection is key. If you don’t care about detecting cars, simply remove that object from the settings.json file and set it to only detect “person."