YOLOv8

约 849 个字 175 行代码 预计阅读时间 5 分钟

yolov8官方github仓库

ultralytics/README.zh-CN.md at main · ultralytics/ultralytics · GitHub

yolov8官方使用文档

Home - Ultralytics YOLOv8 Docs

1

(Windows)RuntimeError解决方案

Windows使用PyTorch遇到RuntimeError: Unable to find a valid cuDNN algorithm to run convolution的解决方案 - 知乎 (zhihu.com)

2

RuntimeError: Dataset 'data\Helmet.yaml' error  
Dataset 'data\Helmet.yaml' images not found , missing paths ['C:\\Users\\Administrator\\datasets\\data\\datasets\\Helmet\\val_list.txt']
Note dataset download directory is 'C:\Users\Administrator\datasets'. You can update this in 'C:\Users\Administrator\AppData\Roaming\Ultralytics\settings.yaml'

修改，

可以改成相对路径

datasets_dir: ..\

3

RuntimeError: 
            Attempt to start a new process before the current process
            has finished its bootstrapping phase.
            This probably means that you are on Windows and you have
            forgotten to use the proper idiom in the main module:
                if __name__ == '__main__':
                    freeze_support()
                    ...
            The "freeze_support()" line can be omitted if the program
            is not going to be frozen to produce a Windows executable.

操作：

model = YOLO("yolov8n.pt")  # 加载预训练模型（建议用于训练）

if __name__ == '__main__': # 添加__main__
    # 使用模型
    model.train(data="data/Helmet.yaml", epochs=3, batch=2[, name="Name"])  # 训练模型

python - YOLOv8 : RuntimeError: An attempt has been made to start a new process before the current process has finished its bootstrapping phase - Stack Overflow

4

MemoryError 调整 batch=

About Memory Error while Training yolov8x-seg · Issue #2916 · ultralytics/ultralytics (github.com)

5

页面大小不够：

python - How to efficiently run multiple Pytorch Processes / Models at once ? Traceback: The paging file is too small for this operation to complete - Stack Overflow

OSError: [WinError 1455] The paging file is too small for this operation to complete. Error loading "...\cusolver64_xx.dll    ...\cudnn_adv_infer64_8.dll" or one of its dependencies.

新建文件 fixNvPe.py

Python Script to disable ASLR and make nv fatbins read-only to reduce memory commit (github.com)

fixNvPe.py

# Simple script to disable ASLR and make .nv_fatb sections read-only
# Requires: pefile  ( python -m pip install pefile )
# Usage:  fixNvPe.py --input path/to/*.dll

import argparse
import pefile
import glob
import os
import shutil

def main(args):
    failures = []
    for file in glob.glob( args.input, recursive=args.recursive ):
        print(f"\n---\nChecking {file}...")
        pe = pefile.PE(file, fast_load=True)
        nvbSect = [ section for section in pe.sections if section.Name.decode().startswith(".nv_fatb")]
        if len(nvbSect) == 1:
            sect = nvbSect[0]
            size = sect.Misc_VirtualSize
            aslr = pe.OPTIONAL_HEADER.IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE
            writable = 0 != ( sect.Characteristics & pefile.SECTION_CHARACTERISTICS['IMAGE_SCN_MEM_WRITE'] )
            print(f"Found NV FatBin! Size: {size/1024/1024:0.2f}MB  ASLR: {aslr}  Writable: {writable}")
            if (writable or aslr) and size > 0:
                print("- Modifying DLL")
                if args.backup:
                    bakFile = f"{file}_bak"
                    print(f"- Backing up [{file}] -> [{bakFile}]")
                    if os.path.exists( bakFile ):
                        print( f"- Warning: Backup file already exists ({bakFile}), not modifying file! Delete the 'bak' to allow modification")
                        failures.append( file )
                        continue
                    try:
                        shutil.copy2( file, bakFile)
                    except Exception as e:
                        print( f"- Failed to create backup! [{str(e)}], not modifying file!")
                        failures.append( file )
                        continue
                # Disable ASLR for DLL, and disable writing for section
                pe.OPTIONAL_HEADER.DllCharacteristics &= ~pefile.DLL_CHARACTERISTICS['IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE']
                sect.Characteristics = sect.Characteristics & ~pefile.SECTION_CHARACTERISTICS['IMAGE_SCN_MEM_WRITE']
                try:
                    newFile = f"{file}_mod"
                    print( f"- Writing modified DLL to [{newFile}]")
                    pe.write( newFile )
                    pe.close()
                    print( f"- Moving modified DLL to [{file}]")
                    os.remove( file )
                    shutil.move( newFile, file )
                except Exception as e:
                    print( f"- Failed to write modified DLL! [{str(e)}]")
                    failures.append( file )
                    continue

    print("\n\nDone!")
    if len(failures) > 0:
        print("***WARNING**** These files needed modification but failed: ")
        for failure in failures:
            print( f" - {failure}")







def parseArgs():
    parser = argparse.ArgumentParser( description="Disable ASLR and make .nv_fatb sections read-only", formatter_class=argparse.ArgumentDefaultsHelpFormatter )
    parser.add_argument('--input', help="Glob to parse", default="*.dll")
    parser.add_argument('--backup', help="Backup modified files", default=True, required=False)
    parser.add_argument('--recursive', '-r', default=False, action='store_true', help="Recurse into subdirectories")

    return parser.parse_args()


###############################
# program entry point
#
if __name__ == "__main__":
    args = parseArgs()
    main( args )

无依赖安装依赖

python -m pip install pefile

运行命令行

python fixNvPe.py --input=E:\Programs\Anaconda3\envs\yolov7\lib\site-packages\torch\lib\*.dll

6

用命令行检测

yolo predict model=yolov8n.pt source='bus.jpg'

代码在：

Python - Ultralytics YOLOv8 Docs

• 调用相机 source="0"
• 检测整个文件夹的图片/视频，source=".../folder"
• 检测单个图片/视频（可以直接写路径，或者用类加载，三种方法）

• show=展示图片
• save=保存图片(到runs/detect/[predictxx]文件夹中)
• name=保存的文件夹的名称

from ultralytics import YOLO
from PIL import Image
import cv2

model = YOLO("model.pt")
# accepts all formats - image/dir/Path/URL/video/PIL/ndarray. 0 for webcam
results = model.predict(source="0")
results = model.predict(source="folder", show=True) # Display preds. Accepts all YOLO predict arguments

# from PIL
im1 = Image.open("bus.jpg")
results = model.predict(source=im1, save=True)  # save plotted images

# from ndarray
im2 = cv2.imread("bus.jpg")
results = model.predict(source=im2, save=True, save_txt=True)  # save predictions as labels

# from list of PIL/ndarray
results = model.predict(source=[im1, im2])

支持的文件类型

Predict - Ultralytics YOLOv8 Docs

参数的作用

Predict - Ultralytics YOLOv8 Docs

7

.yaml文件里面的标签要与labels.txt里面的顺序一致

8

v8 .yaml 文件示例：

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: ../datasets/coco8  # dataset root dir
train: images/train  # train images (relative to 'path') 4 images
val: images/val  # val images (relative to 'path') 4 images
test:  # test images (optional)

# Classes (80 COCO classes)
names:
  0: person
  1: bicycle
  2: car
  ...
  77: teddy bear
  78: hair drier
  79: toothbrush

Object Detection Datasets Overview - Ultralytics YOLOv8 Docs

其中 train: val: 之后的内容有三种写法，

1. 文件夹路径
2. txt文件路径，txt文件内为各个图片的路径
3. 列表形式的图片路径

9

如果detect时使用某个pt模型报错，有可能是因为 训练该模型时使用的ultralytics版本 比 本机安装的版本 新，使用了旧版本 requirements.txt 中未安装的包，因此更新ultralytics包即可

10

train时要修改batch，修改 batch 参数： batch=x

默认为16，-1为自动调整

Configuration - Ultralytics YOLOv8 Docs

11

...
  File "E:\Anaconda3\envs\Pytorch\Lib\site-packages\ultralytics\engine\trainer.py", line 537, in save_metrics
    with open(self.csv, 'a') as f:
         ^^^^^^^^^^^^^^^^^^^
PermissionError: [Errno 13] Permission denied: 'run\\detect\\data1st2\\results.csv'

进程已结束，退出代码为 1

不知什么原因，报错原因应该是 result.csv 文件被打开被占用，但我并没有打开这个文件。

重新再运行程序即可

Training fails when results.csv file is open · Issue #862 · ultralytics/ultralytics (github.com)

12

...
  File "E:\Anaconda3\envs\Pytorch\Lib\site-packages\ultralytics\utils\tal.py", line 152, in get_box_metrics
    bbox_scores[mask_gt] = pd_scores[ind[0], :, ind[1]][mask_gt]  # b, max_num_obj, h*w
                           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
RuntimeError: CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported ar some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.


进程已结束，退出代码为 1

...

大多数情况下，CUDA运行时错误可能是某些索引不匹配的原因，例如您尝试在具有 10 个标签的数据集上训练具有15个输出节点的网络。这个 CUDA 错误的事情是，一旦你收到这个错误一次，你就会在使用 torch.tensors 执行的每个操作中收到它。这会强制您重新启动笔记本。

...

而后经过检查发现，是由于 labels.txt 文件中只有11个标签，而标签文件中有 11 (第12个标签)

python - Pytorch fails with CUDA error: device-side assert triggered on Colab - Stack Overflow

13

恢复/继续之前的训练

Train - Ultralytics YOLOv8 Docs

(可以新建一个 resume_train.py 文件)

from ultralytics import YOLO

# Load a model
model = YOLO('path/to/last.pt')  # load a partially trained model

# Resume training
results = model.train(resume=True)

from ultralytics import YOLO

if __name__ == '__main__':
    # Load a model
    model = YOLO('runs/detect/data1st9/weights/last.pt')  # load a partially trained model

    # Resume training
    results = model.train(resume=True)

14

NotImplementedError: Could not run 'torchvision::nms' with arguments from the 'CPU' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'torchvision::nms' is only available for these backends: [CUDA, Meta, QuantizedCUDA, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, AutogradMPS, AutogradXPU, AutogradHPU, AutogradLazy, AutogradMeta, Tracer, AutocastCPU, AutocastXPU, AutocastCUDA, AutocastPrivateUse1, FuncTorchBatched, BatchedNestedTensor, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PreDispatch, PythonDispatcher].

python - Getting "NotImplementedError: Could not run 'torchvision::nms' with arguments from CUDA backend" despite having all necessary libraries and imports - Stack Overflow

根据最高赞回答，卸载重装torch torchvision torchaudio，再次运行无报错

15

results = model(source=...)

预测后返回的结果，是一个列表，里面的元素的类型是 ultralytics.engine.results.Results (列表应该是对应多张/帧图像的识别结果)，

每个 Results 类中，

• names 属性是以类别序号和类别名字为键值对的字典，如

  {0: 'sunglass', 1: 'hat', 2: 'jacket', 3: 'shirt', 4: 'pants', 5: 'shorts', 6: 'skirt', 7: 'dress', 8: 'bag', 9: 'shoe'}
  

• boxes 属性为 ultralytics.engine.results.Boxes 类，记录着识别的结果，

  Boxes 类中的 cls 属性，记录着从图像上识别出的每个物品对应的类别，类型是 torch.Tensor ，可以通过 .tolist() 方法把它转换成列表，如

  [4.0, 4.0, 9.0, 2.0, 3.0, 9.0, 9.0, 9.0, 4.0, 9.0, 8.0, 9.0]
  

---

Last update: 2024-05-14
Created: 2023-07-11