How to access multiple cameras for live AI model detection: Challenges and Infrastructure Needs?

I need to access 100 CCTV cameras simultaneously for live video detection purposes. I will be deploying a custom AI model to do live detection for all the 100 CCTV camera video feeds. I’m seeking advice on how to set up the cloud/premises infrastructure for this project. What are the potential challenges I might face, and what considerations should I consider to ensure smooth operation?

Details:

  • Scalability: What kind of server or cloud infrastructure would be required to handle the data from 100 cameras?
  • Processing Power: What is the good processing power for real-time video feed processing from 100 cameras?
  • Security: What security measures should be in place to protect the data and the feeds from these cameras?
  • Software/Tools: Are there any recommended software or tools that can help manage multiple camera feeds and perform live detection?

Any link/technical documentation or suggestions would be greatly appreciated!

I have tried live detection with a single IP camera, but it resulted in significant lag issues. I am concerned that scaling up to multiple cameras will exacerbate these problems. I was expecting smoother performance and lower latency.