[mlpack] GSOC 21

[Marcus Edel]

Hello Vishwas,

thanks for getting in touch and interest in the project. The outline looks
good to me, especially VGG16 and VGG19 since they are so widely
used. That said my recommendation is to focus on two or three of the
models you listed, e.g. VGG16, VG19 and InceptionV3, you can always
propose to look into another model at the end if there is time left, but
since testing and documentation, I guess in this case a tutorial would
be useful, take time I would focus on a subset.

I hope that was helpful, also the PR is on my review list.

Thanks,
Marcus

> On 26. Mar 2021, at 13:01, Vishwas Chepuri <chepurivishwas360 at gmail.com> wrote:
> 
> Hello everyone!
> 
> I am Vishwas Chepuri, a sophomore at IIT(BHU), Varanasi, India. I have been getting myself familiar with mlpack for the last couple of months. I wanted to get some opinions regarding my project proposal and kindly help me improve it.
> 
> Idea is to implement the following ready to use models,
> 
> 1) VGG16
> 
> 2) VGG19
> 
> 3) InceptionV3
> 
> 4) ResNet50
> 
> 5) ResNet101
> 
> For each of the above models, I would like to implement a class following the class design in the models repo which includes
> 
> the sketch of model with and without FC layers on top of base architecture, include ImageNet weights for both the models (with and without FC layers), implement preprocessing function which includes preprocessing steps with which the above-included weights are trained, write tests and documentation.
> 
> I have opened a PR #49 (https://github.com/mlpack/models/pull/49 <https://github.com/mlpack/models/pull/49>) for VGG16 and VGG19 models, and I am able to get weights using PyTorch-mlpack Weight Converter (https://github.com/kartikdutt18/mlpack-PyTorch-Weight-Translator <https://github.com/kartikdutt18/mlpack-PyTorch-Weight-Translator>). Hopefully, I will complete implementing and including all the above-mentioned functionalities for these two models before GSOC begins.
> 
> I am excited about this project. Kindly let me know your thoughts on this idea. Thanks for reading.
> 
> Regards,
> 
> Vishwas Chepuri
> 
> GitHub ID: vstark21
> _______________________________________________
> mlpack mailing list
> mlpack at lists.mlpack.org <mailto:mlpack at lists.mlpack.org>
> http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack <http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://knife.lugatgt.org/pipermail/mlpack/attachments/20210327/2b5e313c/attachment.htm>