[jira] [Commented] (SINGA-328) Add VGG models for ImageNet classification

["ASF subversion and git services (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SINGA-328?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16142988#comment-16142988 ] 

ASF subversion and git services commented on SINGA-328:
-------------------------------------------------------

Commit 4fe5271d0a09684ddadfd5feb98333f10ddac0a2 in incubator-singa's branch refs/heads/master from wangwei
[ https://git-wip-us.apache.org/repos/asf?p=incubator-singa.git;h=4fe5271 ]

SINGA-328 - Add VGG models for ImageNet classification

Add vgg models and the parameter conversion script.
The converted parameters are stored on S3.
The download links are providied in the readme file.


> Add VGG models for ImageNet classification
> ------------------------------------------
>
>                 Key: SINGA-328
>                 URL: https://issues.apache.org/jira/browse/SINGA-328
>             Project: Singa
>          Issue Type: New Feature
>            Reporter: wangwei
>            Assignee: wangwei
>
> There are 4 VGG models proposed in the paper https://arxiv.org/abs/1409.1556.
> These 4 models are later updated by adding batch-normalization layers.
> In this ticket, we convert the pre-trained parameters from [pytorch|https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py] into SINGA. Users can then fine-tune the VGG models for their own data using the pre-trained parameters.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)