Facial Recognition Improvements Wanted 2023
Many integrators said they do not have comments due to a lack of experience:
"I don't know, it seems like people are still trying to figure it out. It requires a lot of computing power so the more streamlined we can make that, the better. Either push the compute to the camera, in the cloud or a special appliance that is turn-key."
"None at the moment far, without any real-world experience, it all seems good."
"we haven't done much so can't recommend any improvements."
"We cannot answer this question as we have never used facial recognition."
"Not enough experience so far to provide feedback on improvements."
"At this point, I need more exposure to real-time situations."
"We have not had any experience with facial recognition."
"Having not used any, that's a hard question to answer."
"It needs to mature more before I could comment."
"I don't have any to think of at this time."
Regulation & Privacy Protection
Many integrators mentioned privacy concerns and hope for regulation clarity and privacy protection from facial recognition providers:
"My suggestion for the industry overall is to ensure increased notification to users regarding facial recognition used to accommodate privacy concerns. I suggest an improved ability to notify users that facial recognition is active, and their facial image is being monitored. Also, allow users the ability to opt out of automatic facial recognition and submit to a manual check of their identity (which may require them to wait longer for entry). Finally, I suggest increased published data on the accuracy and false positive rates for various systems."
"Better regulatory understanding. What happens if an EU citizen is on camera? Does the system then need to meet GDPR? What about children? What about inadvertent issues with classifying specific races and gender? What about local and state requirements for the storage and classification of data? These are all things that require a litany of legal work to address."
Some integrators shared that they want to see better facial recognition performance in various environments and various demographic groups. They also hope to see fewer setup requirements:
"I would like to see it get good enough that it can detect well at night. In that case, I might use it, along with video analytics that detects a person, to determine if that is a valid person allowed to be on an exterior property. This property would have a low density of people (2 or 3 at most) for each instance."
"Fewer hardware requirements. Some pushback includes requiring a second server and database, with intense hardware requirements for the processing power needed. Would be nice to use one recording server and analytic server all in one."
"1. Improved accuracy in the night zone (black and white) 2. The face can be recognized even if the number of pixels is small. 3. In general, the accuracy of face recognition is improved."
"We have so many high-resolution options these days, it comes down to having the cameras in the right spots. or having the ability to use varifocal lenses to narrow FOV to allow for longer FR applications."
"Real, onsite performance especially outside and under difficult light/camera positions needs to improve massive."