[SIGCOMM'20] Reducto: On-Camera Filtering For Rersource-Efficient Real-Time Video Analytics

ganler / ResearchReading

General system research material (not limited to paper) reading notes.

GNU General Public License v3.0

20 stars 1 forks source link

[SIGCOMM'20] Reducto: On-Camera Filtering For Rersource-Efficient Real-Time Video Analytics #5

Closed ganler closed 4 years ago

ganler commented 4 years ago

http://web.cs.ucla.edu/~harryxu/papers/li-sigcomm20.pdf

ganler commented 4 years ago

Do frame filtering on the camera-side rather than the server-side.
Commodity cameras have limited compute resources that only permit filtering via frame differencing based on low-level video features, which may lead to unacceptable accuracy loss.

To overcome this, Reducto dynamically adapts filtering decisions according to the time-varying correlation between feature type, filtering threshold, query accuracy, and video content.

ganler commented 4 years ago

Reducto controls the frame rate via tuning the threshold of low-level video features. (Why low-level: because commodity cameras have very limited computing resources that can only cope with simple features.)