0
$\begingroup$

I have trained a CNN model using TF, and now I want to test(evaluate) it on unseen data. How can i do this ?

$\endgroup$

closed as off-topic by Alexis, Juho Kokkala, jbowman, Jeremy Miles, Michael R. Chernick May 31 '18 at 1:41

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "This question appears to be off-topic because EITHER it is not about statistics, machine learning, data analysis, data mining, or data visualization, OR it focuses on programming, debugging, or performing routine operations within a statistical computing platform. If the latter, you could try the support links we maintain." – Juho Kokkala, jbowman, Jeremy Miles, Michael R. Chernick
If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ Is this a question about how to use Tensorflow? If not, could you clarify what the actual question is $\endgroup$ – Juho Kokkala May 30 '18 at 18:13
0
$\begingroup$

When you build the model (including training and testing), you can hold out some data for some final "unseen data test".

In other words, suppose you have 10K data, you can have 4K training, 4K testing, and final 2K for "unseen data test".

Note that, many terms exists and confusing. Such as "validation data set" or "test set", it can mean different things. You may further to research and get details.

$\endgroup$
  • $\begingroup$ Thanks for your answer. Problem is my model is giving me really high accuracy in a small number of iterations which doesn't seem right. So i wanted to somehow provide unseen data, to model and observe its actual perfomence. $\endgroup$ – Salman Shaukat May 30 '18 at 14:28
  • $\begingroup$ Do you think, save/restore could work? as far as data is concerned i have seprated CSV files of newly recorded data which model has never seen, so splitting is not really an issue $\endgroup$ – Salman Shaukat May 30 '18 at 14:30
  • $\begingroup$ @SalmanShaukat many people build model with "fixed amount" of data. That is why those terms exist. If you have real "newly collected / unseen" data, that is even better. $\endgroup$ – Haitao Du May 30 '18 at 14:32

Not the answer you're looking for? Browse other questions tagged or ask your own question.