python write_numbers.py
this takes a sequence of digits as inputs and generates a video as output where the machine generated hand written representation of the digits twine from one to the other sequentially. Though to run this file one has to download the model-4000.h5 file
This application turns a sequence of digits into a video where their handwritten representation twine from one to the other. Each frame of this video comes from an ACGAN generator function.
The above task is carried out by an implementation of an ACGAN(Auxiliary Classifier Generative Adverserial NetWorks). Except for the fact that here we can give ambiguous class intent to the generator. This is acheived through a representtion of the class intent as a vector with the respective probabilities for each class. A custom embedding has been created to learn the representation of this kind which enables us to build the model. The results that come out are quite convincing. One can download the writing_numbers.avi video to see the performance for himself.