I have some queries related to the speed of pretrained networks used for feature extraction.
Query 1:
I am using pre-trained network Googlenet (installed from MATLAB Add-Ons) for feature extraction from images. Right now, I am able to extract features from 5 frames in one second. If I do not have option to enhance the computational power of my machine (on which I am performing feature extraction), how can I increase the speed of feature extraction using the same network (means how can I be able to extract features from, say 20 or 30 frames per second).
Query 2:
What are the different factors that determine the speed of pre-trained networks for feature extraction? Can anyone please elaborate this one?
If my queries are not clear, feel free to comment.I have some queries related to the speed of pretrained networks used for feature extraction.
Query 1:
I am using pre-trained network Googlenet (installed from MATLAB Add-Ons) for feature extraction from images. Right now, I am able to extract features from 5 frames in one second. If I do not have option to enhance the computational power of my machine (on which I am performing feature extraction), how can I increase the speed of feature extraction using the same network (means how can I be able to extract features from, say 20 or 30 frames per second).
Query 2:
What are the different factors that determine the speed of pre-trained networks for feature extraction? Can anyone please elaborate this one?
If my queries are not clear, feel free to comment. I have some queries related to the speed of pretrained networks used for feature extraction.
Query 1:
I am using pre-trained network Googlenet (installed from MATLAB Add-Ons) for feature extraction from images. Right now, I am able to extract features from 5 frames in one second. If I do not have option to enhance the computational power of my machine (on which I am performing feature extraction), how can I increase the speed of feature extraction using the same network (means how can I be able to extract features from, say 20 or 30 frames per second).
Query 2:
What are the different factors that determine the speed of pre-trained networks for feature extraction? Can anyone please elaborate this one?
If my queries are not clear, feel free to comment. deep learning, pre-trained networks, googlenet, machine learning, feature extraction, speed of neural networks MATLAB Answers — New Questions

​