search

JCruan519 / VM-UNet

(ARXIV24) This is the official code repository for "VM-UNet: Vision Mamba UNet for Medical Image Segmentation".
Apache License 2.0
393 stars 12 forks source link

reproduce the exp results #47

Open fAnchao1 opened 2 months ago

fAnchao1 commented 2 months ago

i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.

JAYCHOU2020 commented 2 months ago

i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.

i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.

Do you get 81.3 miou after isic18 training?

fAnchao1 commented 2 months ago

yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.

JAYCHOU2020 commented 2 months ago

yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.

thanks!!!!!! I'll try it

sakurarma commented 2 months ago

@fAnchao1 I meet the same problem. I used the same parameters and fixed seed on another task, but the results were completely different each time. Have you solved it?

fAnchao1 commented 2 months ago

@sakurarma not yet

1362356431 commented 2 months ago

I meet the same problem.I just changed the dataset，but the results were diffident each time on the same GPU.

JAYCHOU2020 commented 2 months ago

yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.

i still cant get the 81.3 miou only 80.5 and my batchsize is 20 when the depth is [2,2,9,2] ,What is your batch size setting? May i see your train.info.log？thanks a lot!

fAnchao1 commented 2 months ago

32，what miou you get

---Original--- From: @.> Date: Sat, May 4, 2024 13:53 PM To: @.>; Cc: @.**@.>; Subject: Re: [JCruan519/VM-UNet] reproduce the exp results (Issue #47)

yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.

i still cant get the 81.3 miou ,What is your batch size setting? May i see your train.info.log？thanks a lot!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

JAYCHOU2020 commented 2 months ago

32，what miou you get … ---Original--- From: @.> Date: Sat, May 4, 2024 13:53 PM To: @.>; Cc: @.**@.>; Subject: Re: [JCruan519/VM-UNet] reproduce the exp results (Issue #47) yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results. i still cant get the 81.3 miou ,What is your batch size setting? May i see your train.info.log？thanks a lot! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

only get miou: 0.8029449451678118