The best choice for your experiment?
ZealACMer opened this issue · 4 comments
ZealACMer commented
In the code, your model choices include ['seqvlad', 'bidirect', 'unshare_bidirect'], which is the choice you made in the experiment to obtain the best accuracy. Thank you very much.
youjiangxu commented
Hi,
Our method uses only the choice of 'seqvlad' on the action recognition
task. In addition, the choices of ['seqvlad', 'bidirect',
'unshare_bidirect'] are used in the task of video captioning and we find
that 'bidrect' could obtain a performance improvement in certain
circumstances.
Ao Li <notifications@github.com> 于2019年5月28日周二 下午3:13写道:
… In the code, your model choices include ['seqvlad', 'bidirect',
'unshare_bidirect'], which is the choice you made in the experiment to
obtain the best accuracy. Thank you very much.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#4?email_source=notifications&email_token=ABVM4BATE2XCBHFF3XYQBTLPXTLTFA5CNFSM4HQAMKP2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GWEXEVQ>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABVM4BGGYQ5HVWMGF45RYI3PXTLTFANCNFSM4HQAMKPQ>
.
ZealACMer commented
youjiangxu commented
Yes, you are right.
Ao Li <notifications@github.com> 于2019年5月28日周二 下午4:52写道:
… Thank you very much for your prompt reply. So for the task of action
recognition, your best choice in the experiments is *the
self.seqvlad_type = 'seqvlad' in Class SeqVLAD*, is that right? Thank you
for your attention.
[image: image]
<https://user-images.githubusercontent.com/16794553/58464768-de706300-8168-11e9-854a-795e200d05e4.png>
[image: image]
<https://user-images.githubusercontent.com/16794553/58464794-e9c38e80-8168-11e9-8842-58b733a82453.png>
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#4?email_source=notifications&email_token=ABVM4BD5JITDOHPCI4OCYZTPXTXDJA5CNFSM4HQAMKP2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWLNY7A#issuecomment-496426108>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABVM4BBLW5CJ3WENPBXPBJ3PXTXDJANCNFSM4HQAMKPQ>
.
ZealACMer commented
Thanks a lot.