Hope can also provide yolov9-s and m model,thanks

[visonpon]

No description provided.

[WongKinYiu]

Currently we plan to release yolov9-s and m models after the paper is accepted and published. If our plan changes, we will directly release the models on the repo.

[xinsuinizhuan]

need to test the yolov9-s model, when to release them?

[huihui308]

Can you give us a time when s model release?

[xinsuinizhuan]

need yolov9-s和yolov9-n model? when release them?

[xinsuinizhuan]

release yolov9s and yolov9n model!!!

[stackByStack]

Sorry, but when could we know whether your models could be released?

[long1109]

Hello Guys, are these released all? Currently i have checked in "https://github.com/WongKinYiu/yolov9/releases/", it is still not released?

[xinsuinizhuan]

still not release t and s model!! when to release them?

[zylo117]

这是什么操作？发布了成绩不公开模型权值，连模型配置、结构都不公开。非要说接收了论文才公开权值。是害怕被偷师改完抢发yolo10吗？要不要看看开源社区对yolov9的支持是怎么样的？是0，我没有看到任何第三方框架宣布对yolov9的支持。正因为作者迟迟不公开细节，人家都不知道怎么复现。
作者既不急着认自己是正统，又害怕别人抢了作者的正统，就挺怪的。

[wenhui-ml]

close yolov9, yolov7-plus will be nice.
Your Team is going ahead just as Close-AI.

[SMRaza1009]

Is there any way to contribute in YOLOv9? For model releasing or any other thing required

[xinsuinizhuan]

Is there any way to contribute in YOLOv9? For model releasing or any other thing required

go to use the yolov10, v10 is better.

[WongKinYiu]

yolov9-s and yolov9-m are released, you could try them.

[long1109]

Thank you very much! Let us try

…

On Wed, Jun 5, 2024 at 6:59 PM Kin-Yiu, Wong ***@***.***> wrote: yolov9-s and yolov9-m are released, you could try them. — Reply to this email directly, view it on GitHub <#3 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANOD72G7UKMVLKLJWASO4FDZF34SPAVCNFSM6AAAAABDUGQMACVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNBZGY2TKNZQG4> . You are receiving this because you commented.Message ID: ***@***.***>

[aaron-hgx]

yolov9-s and yolov9-m are released, you could try them.

Thanks!
Why is there no auxillary branch in YOLOv9-s model? I couldnt find in the config file? Would be really grateful if you could explain.

[WongKinYiu]

yolov9-m and yolov9-s have auxiliary branch.
Provided weights files are reparameterized, which auxiliary branch have removed.

[aaron-hgx]

I understand that the weight files are re-parametrized. However, I do not get why I do not see reversible aux branch in the config file like in the YOLOv9-C config file(which has the comment showing multi-level aux branch part). Can you please explain if i am missing something basic? I would be grateful if you could pinpoint the branches.

YOLOv9

parameters

nc: 80 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 1.0 # layer channel multiple
#activation: nn.LeakyReLU(0.1)
#activation: nn.ReLU()

anchors

anchors: 3

gelan backbone

backbone:
[

conv down

[-1, 1, Conv, [32, 3, 2]], # 0-P1/2

conv down

[-1, 1, Conv, [64, 3, 2]], # 1-P2/4

elan-1 block

[-1, 1, ELAN1, [64, 64, 32]], # 2

avg-conv down

[-1, 1, AConv, [128]], # 3-P3/8

elan-2 block

[-1, 1, RepNCSPELAN4, [128, 128, 64, 3]], # 4

avg-conv down

[-1, 1, AConv, [192]], # 5-P4/16

elan-2 block

[-1, 1, RepNCSPELAN4, [192, 192, 96, 3]], # 6

avg-conv down

[-1, 1, AConv, [256]], # 7-P5/32

elan-2 block

[-1, 1, RepNCSPELAN4, [256, 256, 128, 3]], # 8
]

elan head

head:
[

elan-spp block

[-1, 1, SPPELAN, [256, 128]], # 9

up-concat merge

[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 6], 1, Concat, [1]], # cat backbone P4

elan-2 block

[-1, 1, RepNCSPELAN4, [192, 192, 96, 3]], # 12

up-concat merge

[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P3

elan-2 block

[-1, 1, RepNCSPELAN4, [128, 128, 64, 3]], # 15

avg-conv-down merge

[-1, 1, AConv, [96]],
[[-1, 12], 1, Concat, [1]], # cat head P4

elan-2 block

[-1, 1, RepNCSPELAN4, [192, 192, 96, 3]], # 18 (P4/16-medium)

avg-conv-down merge

[-1, 1, AConv, [128]],
[[-1, 9], 1, Concat, [1]], # cat head P5

elan-2 block

[-1, 1, RepNCSPELAN4, [256, 256, 128, 3]], # 21 (P5/32-large)

elan-spp block

[8, 1, SPPELAN, [256, 128]], # 22

up-concat merge

[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 6], 1, Concat, [1]], # cat backbone P4

elan-2 block

[-1, 1, RepNCSPELAN4, [192, 192, 96, 3]], # 25

up-concat merge

[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P3

elan-2 block

[-1, 1, RepNCSPELAN4, [128, 128, 64, 3]], # 28

detect

[[28, 25, 22, 15, 18, 21], 1, DualDDetect, [nc]], # Detect(P3, P4, P5)
]

[WongKinYiu]

yolov9-m use multi-level reversible aux branch.
yolov9-s use multi-level aux branch.

[felixkarevo]

Do you plan to release a smaller version of YOLOv9 for segmentation tasks? Smaller than yolov9c-seg.

[Zhengcv]

Can you provide the parameter configuration for YOLOv9s on the COCO dataset, especially the key settings for data augmentation such as mixup, scale, and copy-paste? You mentioned the parameters in the paper—are these default parameters applicable to all models?