I was wondering if its possible to develop and train a neural network that generates training data for another network (or possibly itself). I came to this thought wondering the difficulty in creating a labeled dataset for an object detection task - since there is alot to label in case of a multi-class model that maybe needs to be able to detect multiple objects.

So is it possible to automate the process of labeling a dataset for object detection by using an already trained network that spits out the bounding boxes and classes from an input image? Are GANs capable of doing something like this? I couldn't find anything of the sort on the internet.

Thanks alot.

  • $\begingroup$ GANs are used for such tasks. You train your Generator network by using a discriminator network. At the end if training the generator will be able to generate fake observations from a random vector as input. $\endgroup$ – MachineLearner Jul 24 '19 at 23:16

Your Answer

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

Browse other questions tagged or ask your own question.