Can ConvNets be used for real-time object recognition from video feed?



Convolutional neural network are leading type of feed-forward artificial neural network for image recognition. Can they be used for real-time image recognition for videos (frame by frame), or it takes too much processing (assuming they're written in C-like language)?

For example for classification of type of animals based on the training from huge dataset.


Posted 2016-08-07T19:49:25.977

Reputation: 9 163



We are getting there, with as usual some trade-off between quality and speed.

For example Spatial Localization and Detection lecture shows some benchmarks (mAP = Mean Average Precision, higher is better; FPS = frame per second):

Table/Performance: Real-Time Detectors

Table/Performance: R-CNN, Fast R-CNN, Faster R-CNN

Franck Dernoncourt

Posted 2016-08-07T19:49:25.977

Reputation: 1 756