iscyy / yoloair

🔥🔥🔥 专注于YOLOv5，YOLOv7、YOLOv8、YOLOv9改进模型，Support to improve backbone, neck, head, loss, IoU, NMS and other modules🚀

Hi there 👋

Talking about Personal Stuffs:

• 😄Learning direction: Computer vision, algorithm enthusiasts, have some exposure to machine learning, deep learning, software program development;
• 🌟Work direction: software development
• 👯Internship experience: Tencent Ltd., NetEase Research Institute

Languages and Tools

Knowledge Graph

• loading...
• loading...
• loading...

yoloair's Issues

MobileOne如何导出重参化模型?

看你的MobileOne文章 描述训练和推理模型不一样~~
使用 configs/yolov5-Improved/yolov5s_mobileone_backbone.yaml 训练出来的是training模型吗？ 如何转化为推理模型?

val.py doesn't work on its own data

val.py doesn't work on its own data

XCRS源码

请问模块XCRS源码在哪哇 没有找到

RuntimeError: result type Float can't be cast to the desired output type __int64

happen thar error" RuntimeError: result type Float can't be cast to the desired output type __int64" while training?

RuntimeError: shape '[24, 1, 6400]' is invalid for input of size 460800

如何使用GardCAM、GardCAM++等方法进行YOLO系列网络模型热力图可视化

如何使用GardCAM、GardCAM++等方法进行YOLO系列网络模型热力图可视化

Is it possible to adjust the study rate? If so, where can I adjust it?

Is it possible to adjust the study rate? If so, where can I adjust it?？

yolox:AttributeError: 'DetectX' object has no attribute 'anchors'

I use yolox appearing the following error:
Traceback (most recent call last):
File "train.py", line 690, in
main(opt)
File "train.py", line 586, in main
train(opt.hyp, opt, device, callbacks)
File "train.py", line 253, in train
check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
File "/yoloair-main/utils/autoanchor.py", line 43, in check_anchors
anchors = m.anchors.clone() * m.stride.to(m.anchors.device).view(-1, 1, 1) # current anchors
File "/.conda/envs/yolo/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1185, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'DetectX' object has no attribute 'anchors'

Hi, Thank you for your work.I want to know the architecture for this yoloair

运行ACmix报错：RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.cuda.HalfTensor) should be the same

你好，最新版的yoloair在训练acmix的时候报如下错误

Traceback (most recent call last):
File "E:\code\yolov5\yoloair-main\train.py", line 690, in
main(opt)
File "E:\code\yolov5\yoloair-main\train.py", line 586, in main
train(opt.hyp, opt, device, callbacks)
File "E:\code\yolov5\yoloair-main\train.py", line 125, in train
model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
File "E:\code\yolov5\yoloair-main\models\yolo.py", line 132, in init
m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward
File "E:\code\yolov5\yoloair-main\models\yolo.py", line 165, in forward
return self._forward_once(x, profile, visualize) # single-scale inference, train
File "E:\code\yolov5\yoloair-main\models\yolo.py", line 188, in _forward_once
x = m(x) # run
File "D:\Program Files\anaconda3\envs\pytorch\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "E:\code\yolov5\yoloair-main\models\common.py", line 850, in forward
unfold_k = self.unfold(self.pad_att(k_att)).view(b * self.head, self.head_dim,
File "D:\Program Files\anaconda3\envs\pytorch\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "D:\Program Files\anaconda3\envs\pytorch\lib\site-packages\torch\nn\modules\padding.py", line 174, in forward
return F.pad(input, self.padding, 'reflect')
File "D:\Program Files\anaconda3\envs\pytorch\lib\site-packages\torch\nn\functional.py", line 4189, in _pad
return torch._C._nn.reflection_pad2d(input, pad)
RuntimeError: Argument #4: Padding size should be less than the corresponding input dimension, but got: padding (256, 256) at dimension 3 of input 4

Hornet

Thanks for your work, just amazing!

I noticed that there are many models that use Hornet, but there is no corresponding horblock in common.py and yolo.py. Looking forward to further improvement of the project.

By the way, I currently need a model with the best trade-off between mAP and latency. Since I currently don't have enough equipment and time to try all the models, do you have any recommendations?

assert 'nc' in data, "Dataset 'nc' key missing."

遇到这个问题是怎么回事的？

The fusevggforward() is missing

while using yolov5-lite model to val,m.fusevggforward() is missing In /models/Models/Litemodel.py,RepVGGBlock class
Please add missing code

    def fusevggforward(self, x):
        return self.nonlinearity(self.rbr_dense(x))

Training a dataset from voc to yolo format gives a ValueError: not enough values to unpack (expected 2, got 1).

Training a dataset from voc to yolo format gives a ValueError: not enough values to unpack (expected 2, got 1).

训练yolox-s是会出现同样的错误

Traceback (most recent call last):
File "train.py", line 690, in
main(opt)
File "train.py", line 586, in main
train(opt.hyp, opt, device, callbacks)
File "train.py", line 253, in train
check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
File "/home/yoloair-main/utils/autoanchor.py", line 43, in check_anchors
anchors = m.anchors.clone() * m.stride.to(m.anchors.device).view(-1, 1, 1) # current anchors
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 947, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'DetectX' object has no attribute 'anchors'

从头训练yolox模型，报错 'DetectX' object has no attribute 'anchors'

Plotting labels to runs/train/exp4/labels.jpg...
Traceback (most recent call last):
File "/home/innovation/zgx/yoloair-main/train.py", line 695, in
main(opt)
File "/home/innovation/zgx/yoloair-main/train.py", line 591, in main
train(opt.hyp, opt, device, callbacks)
File "/home/innovation/zgx/yoloair-main/train.py", line 254, in train
check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
File "/home/innovation/zgx/yoloair-main/utils/autoanchor.py", line 43, in check_anchors
anchors = m.anchors.clone() * m.stride.to(m.anchors.device).view(-1, 1, 1) # current anchors
File "/home/innovation/anaconda3/envs/yoloair/lib/python3.7/site-packages/torch/nn/modules/module.py", line 779, in getattr
type(self).name, name))
torch.nn.modules.module.ModuleAttributeError: 'DetectX' object has no attribute 'anchors'

键入的命令是：python /home/innovation/zgx/yoloair-main/train.py --data /home/innovation/zgx/yoloair-main/data/VOC.yaml --weights '' --cfg /home/innovation/zgx/yoloair-main/configs/yolox-Improved/yolox_acmix_backbone.yaml --img 640

请问该如何解决？谢谢

训练YOLOX-S

用自己的数据集训练本仓库的YOLOX-S比官方YOLOX-S要高6-7个点，本仓库YOLOX代码和官方YOLOX代码哪里不一样吗？

QueryDet

Are you considering implementing QueryDet(CVPR 2022) network with yolov5 style? I think it's great!
https://github.com/ChenhongyiYang/QueryDet-PyTorch

运行detect.py报错

请问是否有计划提供tensorrt转换工具

YOLOv6模型-v2.0 paper 版本 config

您好，在beta版本中，没看到 YOLOv6模型-v2.0 paper 版本 的config 文件

RuntimeError: expected scalar type Half but found Float

使用yolov5进行训练，train的时候没有问题，训练完一个epoch之后，进行验证的时候会报错误：RuntimeError: expected scalar type Half but found Float，请问是哪里出的问题？使用的v100显卡

怎么拉取最新的代码。训练yolov7-tiny 保持不了权重。

RuntimeError: result type Float can't be cast to the desired output type __int64

佬你好，这个错误咋回事捏？

nms

请问，为啥我每轮的 nms 需要花费五六分钟，然后程序跑一轮就停止了？

KeyError: 'ema'

Traceback (most recent call last):
File "detect.py", line 257, in
main(opt)
File "detect.py", line 252, in main
run(**vars(opt))
KeyError: 'ema'

about yolov7

thank you for your working, I have some question hope you can tell me:

change the --cfg to yolov7, by training, I get the weight best.pt
but when val, it cause a error
........torch.nn.modules.module.ModuleAttributeError: 'IDetect' object has no attribute 'end2end'.....

TypeError: 'NoneType' object is not iterable

运行val.py时报错：
Traceback (most recent call last):
File "D:\yoloair-main\val.py", line 399, in
main(opt)
File "D:\yoloair-main\val.py", line 372, in main
run(**vars(opt))
File "D:\Python\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "D:\yoloair-main\val.py", line 204, in run
loss += compute_loss([x.float() for x in train_out], targets)[1] # box, obj, cls
TypeError: 'NoneType' object is not iterable

ModuleNotFoundError: No module named 'timm'

ModuleNotFoundError: No module named 'timm' 是怎么回事呀博主

请问项目中说使用不同的loss、多种正负样本匹配是如何实现的？

单纯的改结构的话就是直接使用不同的yaml配置文件对吗？

configs中没有看到美团yolov6.yaml的文件

能支持更高效的数据加载方式吗？

Search before asking

• I have searched the YOLOAir issues and found no similar feature requests.

Description

我想使用coco数据重新训练一下，发现数据加载到训练的时间特别长，如果使用lmdb会高效一点吗

Use case

减少训练时间

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

NameError: name 'SwinV2_CSPB' is not defined

Error running yolov5s-swintransformer_v2.yaml.

SwinTransformer等几个transformer模块加在yolo模型里的位置有什么依据吗？

任意地方均可添加随意拼凑？有没有做过实验加在哪里效果最好？

I used YOLOv7 to train my dataset, after dozens of epochs the mAP did not go up, it was always around 0.00005, I did not change the parameters, but I used the code on the YOLOv7 side, the mAP was able to go up normally, why is this?

I used YOLOv7 to train my dataset, after dozens of epochs the mAP did not go up, it was always around 0.00005, I did not change the parameters, but I used the code on the YOLOv7 side, the mAP was able to go up normally, why is this?

This is the result of running the code using yolov7's side

请问关于一些非yolo官方的模型权重参数如何获取

Search before asking

• I have searched the YOLOAir issues and found no similar questions.

Question

例如飞桨的PicoDet可以在yoloair中复现成功，但是飞桨官方的权重文件是.pdparams格式，训练这类模型有办法可以使用到他们的官方权重参数吗。

Additional

No response

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument max in method wrapper_clamp_Tensor)

Swin_V2运行报这个错，想问问要怎么解决呢？

RuntimeError: shape '[4, 5, 8, 8, 8, 64]' is invalid for input of size 696320

i use yolov7-swin config to train voc2007,when model run ro val.run(), run this error，while it training successfully but val
x = x.view(B, H // window_size, window_size, W // window_size, window_size, C) RuntimeError: shape '[4, 5, 8, 8, 8, 64]' is invalid for input of size 696320

请问各个版本的yolo的预训练模型在哪

RuntimeError: shape '[32, 4, 1600]' is invalid for input of size 51200

Replace YOLOx-s backbone with MobileNetV3-small,the above-mentioned problems have occurred,I configured it according to yolov5-mobile3s.yaml,
How do I solve it?

WARNING: NMS time limit 10.0s exceeded

验证阶段非常的慢，是什么原因

This happens when I train my dataset with yolox：ValueError: not enough values to unpack (expected 2, got 1)

 Epoch   gpu_mem       box       obj       cls    labels  img_size
  0/99     1.95G     4.366     8.054     1.459       149       640: 100%|██████████| 14/14 [01:18<00:00,  5.64s/it]                                                                
           Class     Images     Labels          P          R     [email protected] [email protected]:.95:   0%|          | 0/1 [00:02<?, ?it/s]                                                          

Traceback (most recent call last):
File "train.py", line 695, in
main(opt)
File "train.py", line 591, in main
train(opt.hyp, opt, device, callbacks)
File "train.py", line 416, in train
results, maps, _ = val.run(data_dict,
File "/root/miniconda3/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/root/autodl-tmp/yoloair/val.py", line 196, in run
out, train_out = model(im) if training else model(im, augment=augment, val=True) # inference, loss outputs
ValueError: not enough values to unpack (expected 2, got 1)

How to train YOLO v3 and v4?

Hi, I wanna train YOLO v3 and v4, can you provide some related .md files? Thanks a lot.

CBH模块

大佬你好，你的CBH在yaml里面是不是配置错了啊

Swin tranformer V2模块报错，请问有解决方案吗?

Type error:roll():argument 'shift' must be tuple of ints,not tuple

Running val.py brings up ValueError: not enough values to unpack (expected 2, got 1)

Running val.py brings up ValueError: not enough values to unpack (expected 2, got 1)

val.py No parser.add_argument('--otaloss', type=str, default='origin', help='use yolov7 or yolox')
How can I solve it?

ModuleNotFoundError: No module named 'timm'

Following setup steps in README.md, when running training script (python train.py --data coco128.yaml --cfg configs/yolov5/yolov5s.yaml), met the following errors:

Downloading https://ultralytics.com/assets/Arial.ttf to /home/logan/.config/Ultralytics/Arial.ttf...
Traceback (most recent call last):
File "/home/logan/dev/pyfarm/pytorch/yoloair/train.py", line 41, in
import val # for end-of-epoch mAP
File "/home/logan/dev/pyfarm/pytorch/yoloair/val.py", line 38, in
from models.common import DetectMultiBackend
File "/home/logan/dev/pyfarm/pytorch/yoloair/models/common.py", line 35, in
from models.Models.muitlbackbone import conv_bn_hswish, MobileNetV3_InvertedResidual, DepthSepConv,
File "/home/logan/dev/pyfarm/pytorch/yoloair/models/Models/muitlbackbone.py", line 26, in
from models.Models.SwinTransformer import SwinTransformerLayer
File "/home/logan/dev/pyfarm/pytorch/yoloair/models/Models/SwinTransformer.py", line 9, in
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
ModuleNotFoundError: No module named 'timm'

Any help?

P.S. My env: Ubuntu22.04.1, python 3.10.4

decoupled head

When I try to use the decoupled head ,I find the Gflops is too large, up to 56.4 GFLOPs,Is this normal?

代码训练问题！

Search before asking

• I have searched the YOLOAir issues and found no similar questions.

Question

请问改这些模块，博主用数据集自己训练过吗？最好可以放上一些有说服力的图、表！！！

Additional

No response