BestYOLO是一个以科研和竞赛为导向的最好的YOLO实践框架!

目前BestYOLO是一个完全基于YOLOv5 v7.0 进行改进的开源库，该库将始终秉持以落地应用为导向，以轻便化使用为宗旨，简化各种模块的改进。目前已经集成了基于torchvision.models 模型为Backbone的YOLOv5目标检测算法，同时也将逐渐开源更多YOLOv5应用程序。

🌟改进

Backbone-ResNet18 对齐 resnet18
Backbone-RegNet_y_400mf 对齐 regnet_y_400mf
Backbone-MobileNetV3 small 对齐 mobilenet_v3_small
Backbone-EfficientNet_B0 对齐 efficientnet_b0
Backbone-ResNet34 对齐 resnet34
Backbone-ResNet50 对齐 resnet50
Backbone-EfficientNetV2_s 对齐 efficientnet_v2_s
Backbone-EfficientNet_B1 对齐 efficientnet_b1
Backbone-MobileNetV2 对齐 mobilenet_v2
Backbone-wide_resnet50_2 对齐 wide_resnet50_2
Backbone-VGG11_BN 对齐 vgg11_bn
Backbone-Convnext Tiny 对齐 convnext_tiny

所有Backbone都支持开启预训练权重，只需添加pretrained=True到每个common.py 的模型中。torchvision.models中的预训练权重都是基于ImageNet-1K数据集训练的！

models	layers	parameters	model size(MB)
yolov5n	214	1766623	3.9
MobileNetV3s	313	2137311	4.7
efficientnet_b0	443	6241531	13.0
RegNety400	450	5000191	10.5
ResNet18	177	12352447	25.1
ResNet34	223	22460607	45.3
ResNet50	258	27560895	55.7
EfficientNetV2_s	820	22419151	45.8
efficientnet_b1	539	6595615	13.8
mobilenet_v2	320	4455295	9.4
wide_resnet50_2	258	70887103	142.3
vgg11_bn	140	10442879	21.9
convnext_tiny	308	29310175	59.0

.yaml配置文件中的depth_multiple和width_multiple可以同时设置为1试试，说不定会有不错的效果。

SPP是空间金字塔池化，作用是一个实现一个自适应尺寸的输出。（传统的池化层如最大池化、平均池化的输出大小是和输入大小挂钩的，但是我们最后做全连接层实现分类的时候需要指定全连接的输入，所以我们需要一种方法让神经网络在某层得到一个固定维度的输出，而且这种方法最好不是resize（resize会失真），由此SPP应运而生，其最早是何凯明提出，应用于RCNN模型）。当今的SPP在faster-rcnn上已经发展为今天的Multi-Scale-ROI-Align，而在Yolo上发展为SPPF。

models	layers	parameters
yolov5n(SPPF)	214	1766623
yolov5n-SPP	217	1766623
yolov5n-SimSPPF	216	1766623
yolov5n-ASPP	214	3831775
yolov5n-RFB	251	1932287
yolov5n-SPPCSPC	232	3375071
yolov5n-SPPCSPC_group	232	2047967
yolov5n-SimCSPSPPF	229	3375071

models	layers	parameters
yolov5n	214	1766623
yolov5n-FPN-AC	188	1858399
yolov5n-PAN-AC	186	1642591
yolov5n-FPN+PAN-AC	160	1734367
yolov5n-FPN-AS	204	2106847
yolov5n-PAN-AS	194	1891039
yolov5n-FPN+PAN-AS	184	2231263

训练不要使用Soft-NMS，耗时太久，请在val阶段开启，适用于小目标重叠数据。

💻应用

YOLOv5微信小程序

YOLOv5数据集制作助手

YOLOv5单框多标签

Fogging Augment

🌈技巧

小目标检测大杀器：yolov5-pip 和sahi
小目标检测大杀器：填鸭式数据增强

应注意：训练和推理数据保持相同的数据形式，即不能通过非切图训练，根据切图推理！

一个不成熟的优化器选择参考：

📋参考

💫工作

🔔引用

@article{2023bestyolo,
  title={{BestYOLO}: Making research and competition easier},
  author={Rongsheng Wang},
  repo={github https://github.com/WangRongsheng/BestYOLO},
  year={2023}
}

🚀贡献

AttributeError: 'EfficientNet' object has no attribute 'yaml'

@WangRongsheng 你好, I used to your repo and i want to perform a training using my own dataset. I want to use efficientNet instead of darknet in my backbone network. Within the train.py file, I replaced my weights with efficientnet_b0.pt instead of yolov5n.pt. The error mentioned in the title appears. I would be happy if you

runfile('D:/BestYOLO/BestYOLO-main/train.py', wdir='D:/BestYOLO/BestYOLO-main')
Reloaded modules: models, utils, utils.downloads, utils.metrics, utils.general, utils.augmentations, utils.torch_utils, utils.dataloaders, utils.segment, utils.segment.general, utils.plots, models.common, classify.val, models.experimental, utils.autoanchor, models.yolo, utils.loggers.clearml, utils.loggers.clearml.clearml_utils, utils.loggers.wandb, utils.loggers.wandb.wandb_utils, utils.loggers
train: weights=efficientnet_b0.pt, cfg=, data=data\coco128.yaml, hyp=data\hyps\hyp.scratch-low.yaml, epochs=2, batch_size=2, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, noplots=False, evolve=None, bucket=, cache=None, image_weights=False, device=, multi_scale=False, single_cls=False, optimizer=SGD, sync_bn=False, workers=8, project=runs\train, name=exp, exist_ok=False, quad=False, cos_lr=False, label_smoothing=0.0, patience=100, freeze=[0], save_period=-1, seed=3407, local_rank=-1, loss_OTA=False, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest
YOLOv5 2023-10-30 Python-3.10.13 torch-2.1.0+cpu CPU

hyperparameters: lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0
ClearML: run 'pip install clearml' to automatically track, visualize and remotely train YOLOv5 in ClearML
Comet: run 'pip install comet_ml' to automatically track and visualize YOLOv5 runs in Comet
TensorBoard: Start with 'tensorboard --logdir runs\train', view at http://localhost:6006/
github: skipping check (not a git repository), for updates see https://github.com/ultralytics/yolov5
Traceback (most recent call last):

File ~.conda\envs\thesis_two\lib\site-packages\spyder_kernels\py3compat.py:356 in compat_exec
exec(code, globals, locals)

File d:\bestyolo\bestyolo-main\train.py:640
main(opt)

File d:\bestyolo\bestyolo-main\train.py:534 in main
train(opt.hyp, opt, device, callbacks)

File d:\bestyolo\bestyolo-main\train.py:125 in train
model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create

File ~.conda\envs\thesis_two\lib\site-packages\torch\nn\modules\module.py:1695 in getattr
raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'")

AttributeError: 'EfficientNet' object has no attribute 'yaml'

wangrongsheng / bestyolo Goto Github PK

bestyolo's Introduction

🌟改进

💻应用

🌈技巧

📋参考

💫工作

🔔引用

🚀贡献

bestyolo's People

Contributors

Stargazers

Watchers

Forkers

bestyolo's Issues

Recommend Projects

Recommend Topics

Recommend Org