MNIST is a famous data set of hand written digits. Suppose we knew who wrote digits for example, female, left handed, 25 yrs old.

How would I use these information in CNN in tensorflow? or any other library.

Digits are images and that what CNN handles well but gender, dominant hand, and age are not images. How would you use that information?


Convolutional layers are useful for images because they take into consideration the neighborhood of pixels. However, for labels like gender and handedness a convolutional layer may not be particularly useful.

However, after the convolutional layers you usually tend to place some densely connected layers. It is there that you may want to add these additional features. When you reshape the 2D matrices which results from the convolutions, you can concatenate the additional features, then feed this new vector to your Dense layer.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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