search

xuebinqin / BASNet

Code for CVPR 2019 paper. BASNet: Boundary-Aware Salient Object Detection
MIT License
1.35k stars 249 forks source link

Large parameters #11

Closed InstantWindy closed 4 years ago

InstantWindy commented 4 years ago

Hi. This model of GPU memory is large. Have you tried to optimize it? Thanks

xuebinqin commented 4 years ago

Thanks for your interests. Unfortunately, we didn't optimize it. We did realize the problem. Instead of optimizing it, we actually design another model which is much smaller than this one. We just submitted our new model to a conference and will release that later.

You could try to optimize it by simply using less filters of certain layers. According to our experience, that wouldn't impact the final performance that much.

On Mon, Sep 2, 2019 at 9:30 PM InstantWindy notifications@github.com wrote:

Hi. This model of GPU memory is large. Have you tried to optimize it? Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NathanUA/BASNet/issues/11?email_source=notifications&email_token=ADSGORNAZ4EEDAZLIYXOJWLQHXK5BA5CNFSM4ITBSQKKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HI4AMAQ, or mute the thread https://github.com/notifications/unsubscribe-auth/ADSGORPQVC2SICGAKMRIOODQHXK5BANCNFSM4ITBSQKA .

-- Xuebin Qin PhD Candidate Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/

InstantWindy commented 4 years ago

I will try. When the model converges, How much has loss fallen?

xuebinqin commented 4 years ago

the summation of the eight losses including the final output and side outputs are around 0.9 and the final output loss is about 0.07.

On Mon, Sep 2, 2019 at 10:06 PM InstantWindy notifications@github.com wrote:

I will try. When the model converges, How much has loss fallen?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/NathanUA/BASNet/issues/11?email_source=notifications&email_token=ADSGORKNMLN5ZEKTLVVDSYLQHXPFLA5CNFSM4ITBSQKKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5W6MLI#issuecomment-527296045, or mute the thread https://github.com/notifications/unsubscribe-auth/ADSGOROQC2QZFRH53BZEXH3QHXPFLANCNFSM4ITBSQKA .

-- Xuebin Qin PhD Candidate Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/