Query about speed of inference on semantic segmentation.

[zwbx]

Hi, thanks for your nick work.
I find it is a little bit slow to inference on semantic segmentation code, the speed is less than 1 task/s on 1 v100.
Is this a normal case? or there are anything to do?
Thanks:-)

[daeunni]

Same here

[daeunni]

@zwbx Hi, did you figure out what is the most reason?
I think this is because CoTTA uses augmentation processes like [0.5, 1.0, 1.5, ..] when doing each image's inferences.

[zwbx]

yes, there is nothing wrong.

…

________________________________ 发件人: Daeun Lee ***@***.***> 发送时间: 2023年11月26日 3:07 收件人: qinenergy/cotta ***@***.***> 抄送: Wenbo Zhang ***@***.***>; Mention ***@***.***> 主题: Re: [qinenergy/cotta] Query about speed of inference on semantic segmentation. (Issue #19) <https://github.com/zwbx> CAUTION: External email. Only click on links or open attachments from trusted senders.

________________________________ @zwbx Hi, did you figure out what is the most reason? I think this is because CoTTA uses augmentation processes like [0.5, 1.0, 1.5, ..] when doing each image's inferences. — Reply to this email directly, view it on GitHub<#19 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AJFDSL4IFYBPJQ5LVN7AMCLYGINGHAVCNFSM6AAAAAAV4YGJ6WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMRWGM3TEOBUGQ>. You are receiving this because you were mentioned.Message ID: ***@***.***>