Mastodawn

OCTranspo wants to use AI to monitor its security cameras.

TIL I learned the STM monitors its security cameras using AI

https://www.cbc.ca/news/canada/ottawa/oc-transpo-ottawa-ai-security-camera-transit-9.6990550
- - -
OCTranspo veut utiliser l'IA pour surveiller ses caméras de sécurité.

Aujourd’hui j’ai appris que la STM surveille ses caméras de sécurité à l'aide de l'IA

// Article en anglais //

#Ottawa #OCtranspo #AI #IA #ArtificialIntelligence #IntelligenceArtificielle

AI could watch Ottawa's transit cameras for crime, emergencies | CBC News

OC Transpo is looking to use artificial intelligence (AI) to alert special constables to crime and emergencies occurring within the agency's vehicles and facilities, Ottawa city councillors have heard.

CBC

Show thread

The Book of Kels Nov 25

@EdwinG

Why is this city so determined to choose the worst "solution" to any problem?

When London tried this, fully *half* of the alerts were false alarms. And other places not any better.

In the sciences, LLMs are good at sorting large amounts of data, but a vital component is expert humans checking the results as part of the process. Guaranteed the city isn't gonna bother with that step, they want a "set it and forget it" solution that doesn't exist.

Show thread

Charles Nov 25

@Nezchan @EdwinG the article doesn't mention using LLM fornthat. No product has been chosen, its in preliminary exploration phase.

Given I'm sure humans are not currently looking at the 10k camera feeds constantly, anomaly detection can probably help human review. All depends on implementation...

Show thread

Edwin G.

Nov 25

@drpyser @Nezchan It could very well be some form of neural network-based image pattern recognition

Show thread

Charles

@EdwinG @Nezchan for sure that's what it would be. Different from LLMs though.

For automatic image processing and "anomaly detection" i don't know of a better (more effective/developed) approach than machine learning and deep neural networks. That's the state of the art as far as I know. False positives are always there, of course, and humans need to review and be critical of the results.