|
|
|
|
|
by nitrogen
3797 days ago
|
|
It's not just feasible, it's existed for decades and is used by security cameras that record to (usually on-premises) DVRs. There have been open source apps that do basically the same thing (detect motion, save video around events, etc.). With modern mobile GPUs you have ample processing power for running over-the-top machine learning algorithms. There is no technical need to violate privacy by default. |
|