Avigilon View - Manual (Page 80)

NOTE: When the camera is moved and its field of view changes, the learning results can no longer be used as
they can no longer accurately identify objects. You should either delete the learning results and create a new set
of results, or use the Factory Default settings. To do so, complete the procedures in Teach By Example History
on the facing page.
Learning From Collected Training Data
Once you have collected training data, you can have the analytic appliance learn from this data, and the results
of the data can be uploaded to the camera. The more data you collect and use to teach, the more accurate the
analytics become. It is also important to have a balance of both true alarms and false alarms.
NOTE: You must mark a minimum of 30 true alarm events and 30 false alarm events for a camera before you can
create a learning file.
1. Click the Teach tool.
2. Click Learn.
3. Once the learning results have been created, the graphic display of alarm events will change to show the
difference in missed and corrected alarms if you uploaded the learning results to a camera.
4. Use the drop-down menu to see the impact of changing the sensitivity of your rules:
l
Use actual rule sensitivity: shows the status of the alarms if no rule sensitivity changes are made.
l
Use What-If sensitivity: shows how the number and status of alarms change if the sensitivity for all
rules violated in a set of alarm events is increased or decreased.
Increasing or decreasing sensitivity with the Use What-if sensitivity does not change the sensitivity
of a rule. You must do this manually for each rule.
Uploading Learning Results to a Camera
After the learning results have been created you can upload them to the camera.
NOTE: When uploading learning results to a camera connected to an analytic appliance, the set of results are
used only on the camera channel where the results were uploaded, not on all channels.
80 Learning From Collected Training Data