I'm working on an image segmentation algorithm with FCN (Long et al., 2015) as the backbone network.

One idea I have is to use the argmax binary mask obtained from the final score layer (250x250x1) to generate some data (e.g. number of blobs in the mask) to modify the ground truth (e.g. set some pixels in the gt mask to 'ignore' labels) or in some way (partly) extract from the features (similar to RPN layer in FasterRCNN).

Does this violate any deep learning or machine learning rules?

  • $\begingroup$ Are you making this modification to labels in the training dataset only (as part of training), or are you also modifying the test dataset labels? $\endgroup$ Commented Feb 8, 2018 at 16:41
  • $\begingroup$ No just the training. Test and validation data remain unchanged. $\endgroup$
    – Alex
    Commented Feb 8, 2018 at 21:58
  • $\begingroup$ So it's modifying the valie of the loss function. Nothing else. $\endgroup$
    – Alex
    Commented Feb 9, 2018 at 13:39

1 Answer 1


No - sounds like you are just stacking different neural networks. Neural networks are inherently stacked models. Sometimes those stacked models are homogeneous and sometimes those stack are heterogeneous.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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