[mlpack] GSOC-2020 | Idea Proposal | Image Generation Modules

[rakesh acharya]

Hello Marcus/Ryan,

I'm Rakesh Acharya , I wish to contribute to* mlpack library* by developing
the *Image Generation* modules.Presently mlpack has *GAN,WGAN*.Many latest
GAN models have disrupted the Image Generation field like BEGAN,Energy
Based GAN,Manifold Matching GAN,Probabilistic GAN.

 *Yann LeCun described it as “the most interesting idea in the last 10
years in Machine Learning”.  *

Among the existing models *BEGAN/I-BEGAN* outruns most of the
models,including Energy Based GANs.I hope it will be a good leap forward in
building mlpack Image Generation modules.

After my initial proposal* Marcus suggested me I-BEGAN* paper and after
giving a thorough reading came to a conclusion that , though it has problem
with disentagling factors  results obtained are *very good with  minimal
Wassertian Distance based losses* and *maxmized mutual information.*

*BEGAN :  <goog_1053210462>*https://arxiv.org/abs/1703.10717
*IBEGAN* : https://www.hindawi.com/journals/cin/2018/6465949/

I'd like to hear from you guys *suggestions regarding adding these
modules to mlpack library* and growing the areas it can be used to *Image
Generation.*

Regards,

Rakesh Acharya .D

Computer Engineering,VIT Chennai
Deep Learning,Computer Vision,Digital Signal Processing
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://knife.lugatgt.org/pipermail/mlpack/attachments/20200315/ec0bf583/attachment.htm>