after the completion of offline training, how can I measure the testing time for individual sample?

I have trained my ANN for fault analysis. Now i am testing it for new data and its working accurately.

But I want to measure the detection time which ANN is taking to identify the new sample???

Answer 1

[Madhav Thakker]

Hi Ali,

You can use tic toc function to calculate the time it takes for each sample during testing. Call tic before the prediction and toc just after prediction. Make sure not to do any other processes like reading or resizing in between these commands.

Hope this helps.

[ali mustafa]

I used this option to find the detection time for new sample but it is not very precise.

If i measure the single sample or more than hundred sample , this command almost gives the detection time which ha not remarkable difference from the single sample detection time

Is there any other way to find the detection time?.

[Madhav Thakker]

Hi, I tried tic toc with different examples (alexnet classficiation, shallow neural network) to replicate your issue but it works as expected. The time increases as I increase the number of testing samples. Can you give more information so that I can replicate your issue?

[ali mustafa]

Dear Sir,

If I test the single sample, the minimum observed detection time is 40ms.

When i increase the number of sample as I tried 772 sample together, and the minimum detection was 70ms.

If its working correctly, it must be much higher than 70ms.

For example, if i calculate the time for one sample among the 772 samples detection time, the time would be very small in proportion as explained below.

1:x=772:70

but for indidual sample, the tic toc command gives more than 40ms. Thats my problem.

Because I am working on fault detection of transmission line, and the detection time is very less like less than 5 ms is found in literature. So, how could I measure that time