Troubleshooting
There are typically two causes for this.
To determine if this is the cause look at the logs for the trigger container in Docker and see if there are any lines like this:
2020-06-15T12:18:21-07:00 [Trigger Dog detector] /images/DogSD.20200615_121725679.jpg: Analyzing
If you do not see lines like the above it means the images aren't getting picked up by the trigger
engine. Double-check that you edited the watchPattern property for the trigger correctly
and that its path and wildcard filename are correct.
To determine if this is the look at the logs for the trigger container in Docker
and see if there are follow-on log messages after Analyzing that look similar to this:
2020-06-15T12:18:22-07:00 [Trigger Dog detector] /images/DogSD.20200615_121725679.jpg: Found at least one object in the photo.
If there are never any log messages other than the one indicating Analyzing then it means Deepstack
is receiving the request to analyze properly but doesn't responds with a result. To fix this change
the Docker image used for Deepstack from deepquestai/deepstack:latest to deepquestai/deepstack:noavx.
You will need to get an API key to use the noavx version of Deepstack. After starting the Docker
images use your browser to access port 5000 on the machine running the image. The Deepstack landing
page will appear and walk through the process of obtaining the API key.
This is typically due to a large number of files in the aiinput folder. It's important to keep the folder clean of older files for good performance. If using BlueIris go to the Clips and archiving tab in settings and change the archiving rules for aiinput to delete at 1 GB and 1 hour.
You can also adjust the sensitivity of each camera in BlueIris to reduce the number of candidate images that get produced. It takes some trial and error to find the right balance here, but in general you don't want the candidate image generation to be too sensitive as it results in a large number of images to process that aren't necessary.