1.服务器Ubuntu系统编程环境安装指令

0.cuda版本切换

gedit /etc/profile

export PATH=$PATH:/usr/local/cuda-11.8/bin

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.8/lib64

export PATH=/usr/local/cuda-11.8/bin${PATH:+:${PATH}}

export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

export PATH=/usr/local/cuda-11.8/bin:$PATH

export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64:$LD_LIBRARY_PATH

/usr/local/MATLAB/R2024a/bin/matlab

export PATH=/usr/local/MATLAB/R2024a/bin/matlab:$PATH

find / -type

d -name "MATLAB_R2024a"

2>/dev/null

一:pytorch___version__

import torch

print(torch.cuda.is_available())

print(torch.cuda.get_device_name())

print(torch.cuda.current_device())

print(torch.cuda.device_count())

print(torch.__version__)

print(torch.version.cuda)

二TZB天智杯4+基本的数据处理

#export PATH=/home/zkti/0.Test_tang/miniconda3/bin:$PATH

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n tzb_4 python=3.8 -y #注意环境命名必须用小写，大写用pycharm调试不通过

source activate tzb_4

conda search gdal

(2.2.4,3.0.2)

conda install gdal==3.0.2

pip install pillow matplotlib scikit-image opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pillow -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install matplotlib -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install scikit-image -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install tqdm -i https://pypi.tuna.tsinghua.edu.cn/simple

三、YOLOv5

1)模型环境配置

https://github.com/ultralytics/yolov5 这里选择yolov5-4.0版本

export PATH=/home/zkti/0.Test_tang/miniconda3/bin:$PATH

conda create -n yolov5 python=3.8 -y

source activate yolov5

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

注意点是：在这里先去掉

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

2）自定义数据集

Yolo官网格式：https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data

pascalvoc转yolo格式

程序在utrla_voc2yolo文件夹里面

1. voc_labelname.py 统计数据库的类别

2. voc2txt,py 将XML文件生成对于txt文件，这里包含坐标转换

3. makedata.py 根据txt文件来移动图像，生成yolo数据集

4. 坐标转换：https://blog.csdn.net/angelbeats11/article/details/88427359

coco转yolo格式

3）训练pascalvoc格式

1)新建文件夹

在yolov5-master内新建predict和weights文件夹

predict：用于存储测试后的图像和TXT

weights:存放预训练模型

2)修改模型参数

Yolov5-master\models\ yolov5l.yaml: nc: 80 # number of classes

/yolov5-4.0/data/hyp.scratch.yaml #从头训练，不借助预训练数据集

yolov5-4.0/data/hyp.finetune.yaml #借助预训练训练，这里可以对超参数修改

3)修改python程序

Yolov5-master\data\ VOC.yaml:

注意点：这里names一定要与voc2txt.py 保持一致

# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/

# Train command: python train.py --data VOC.yaml

# Default dataset location is next to YOLOv5:

# /parent

# /datasets/VOC

# /yolov5

# 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: /home/zkti/data_4T/1.datasets/2.Optical/0.DIOR/DIOR_yolov5/

train: # train images (relative to 'path') 16551 images

- /home/zkti/data_4T/1.datasets/2.Optical/0.DIOR/DIOR_yolov5/images/train

val: # val images (relative to 'path') 4952 images

- /home/zkti/data_4T/1.datasets/2.Optical/0.DIOR/DIOR_yolov5/images/val

test: # test images (optional)

- /home/zkti/data_4T/1.datasets/2.Optical/0.DIOR/DIOR_yolov5/images/test

# Classes

nc: 20 # number of classes

# class names

names: ['windmill', 'stadium', 'overpass', 'ship', 'Expressway-Service-area', 'trainstation', 'storagetank', 'airport', 'tenniscourt', 'chimney', 'groundtrackfield', 'vehicle', 'golffield', 'bridge', 'dam', 'basketballcourt', 'Expressway-toll-station', 'harbor', 'airplane', 'baseballfield']

4)运行训练

(1)配置train.py的模型和数据集格式

#指定GPU训练

import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

#修改parser配置文件

if __name__ == '__main__':

parser = argparse.ArgumentParser()

parser.add_argument('--weights', type=str, default='yolov5s.pt', help='initial weights path') #修改初始化模型位置

parser.add_argument('--cfg', type=str, default='models/yolov5s.yaml', help='model.yaml path') #修改模型配置文件参数的路径

parser.add_argument('--data', type=str, default='data/VOC_DIOR.yaml', help='data.yaml path') #修改数据集路径

parser.add_argument('--hyp', type=str, default='data/hyp.finetune.yaml', help='hyperparameters path') #一定要配置借助预训练模型的超参数

parser.add_argument('--epochs', type=int, default=300)

parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs')

parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='[train, test] image sizes')

parser.add_argument('--rect', action='store_true', help='rectangular training')

parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')

parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')

parser.add_argument('--notest', action='store_true', help='only test final epoch')

parser.add_argument('--noautoanchor', action='store_true', help='disable autoanchor check')

parser.add_argument('--evolve', action='store_true', help='evolve hyperparameters')

parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')

parser.add_argument('--cache-images', action='store_true', help='cache images for faster training')

parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')

parser.add_argument('--device', default='0', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')

parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')

parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')

parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')

parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')

parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')

parser.add_argument('--log-imgs', type=int, default=16, help='number of images for W&B logging, max 100')

parser.add_argument('--log-artifacts', action='store_true', help='log artifacts, i.e. final trained model')

parser.add_argument('--workers', type=int, default=8, help='maximum number of dataloader workers')

parser.add_argument('--project', default='runs/train', help='save to project/name')

parser.add_argument('--name', default='exp', help='save to project/name')

parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')

parser.add_argument('--quad', action='store_true', help='quad dataloader')

opt = parser.parse_args()

(3) 指定GPU:

import os

#多块使用逗号隔开

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

(4)python tran.py

5)运行测试

运行前需要将DIOR_classes的类型顺序要与之前保持一致

Python detect_tang_v5.py

输出格式：类别+置信度+坐标（xmin,ymin,xmax,ymax）

6）得出map

(1) Python detect_tang_v5.py 这里要注意，一定要设置置信度为0.01，具体看mAP

(2) map_tang.py :

超参数：这里一定要这样设置，否则计算map极其慢

parser.add_argument('-na', '--no-animation',default=True, help="no animation is shown.", action="store_true")

parser.add_argument('-np', '--no-plot',default=False, help="no plot is shown.", action="store_true")

四、YOLOv8

参考网址

yolov8测试图片并保存预测结果，用于map的计算

https://blog.csdn.net/qq_42178122/article/details/133688916

yolov8的predict使用方法，更改predict.py的输出结果，输出label的真实坐标，保存图片和txt文档，图片中没有异物生成空的txt文档

https://blog.csdn.net/m0_57847261/article/details/134319876?spm=1001.2101.3001.6650.2&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7ECtr-2-134319876-blog-133688916.235%5Ev43%5Epc_blog_bottom_relevance_base7&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7ECtr-2-134319876-blog-133688916.235%5Ev43%5Epc_blog_bottom_relevance_base7&utm_relevant_index=5

YOLOv8的训练、验证、预测及导出[目标检测实践篇]

https://blog.csdn.net/weixin_45144684/article/details/138454441

目标检测项目实战——使用Yolo v8识别图片中验证拼图位置

https://zhuanlan.zhihu.com/p/667220840

YOLOv8打印检测框坐标

https://blog.csdn.net/GuoKe_L/article/details/136903386?spm=1001.2101.3001.6650.7&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7EBlogCommendFromBaidu%7ERate-7-136903386-blog-139767590.235%5Ev43%5Epc_blog_bottom_relevance_base7&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7EBlogCommendFromBaidu%7ERate-7-136903386-blog-139767590.235%5Ev43%5Epc_blog_bottom_relevance_base7&utm_relevant_index=12

YOLOv8目标检测结果分析

https://blog.csdn.net/weixin_38566632/article/details/140690304

0.模型环境配置完整版-服务器

https://github.com/ultralytics/ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n yolov8 python=3.8 -y

source activate yolov8

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

pip install ultralytics -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pycuda -i https://pypi.tuna.tsinghua.edu.cn/simple

1. 模型环境配置

https://github.com/ultralytics/ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n yolov8 python=3.8 -y

source activate yolov8

conda install pytorch==2.2.1 torchvision==0.17.1 pytorch-cuda=12.1 -c pytorch -c nvidia

pip install ultralytics -i https://pypi.tuna.tsinghua.edu.cn/simple

##########TensorRT###############

pip install tensorrt -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnx -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxsim -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxslim -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxruntime -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pycuda -i https://pypi.tuna.tsinghua.edu.cn/simple

###################SAHI################

pip install opencv-python sahi tabulate podm tqdm argparse -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install object_detection_metrics -i https://pypi.tuna.tsinghua.edu.cn/simple

2. 自定义数据集

数据集格式参考网址为：https://docs.ultralytics.com/datasets/detect/，介绍了Ultralytics YOLO 格式，该格式是为了ultralytics旗下的yolo模型标准数据集格式，使用其他数据集训练时，必须得进行数据转换

1）VOC数据集转Ultralytics YOLO格式

可以借鉴转Ultralytics 给的官方转换方法：2023-Yolov8/ultralytics-main/ultralytics/cfg/datasets/VOC.yaml

使用/home/zkti/0.Test_tang/0.Python_Script/1.HBB/0.Coordinate transformation/utrla_voc2yolo文件下的python脚本

1. voc_labelname.py 统计数据库的类别

2. voc2txt,py 将XML文件生成对于txt文件，这里包含坐标转换

3. makedata.py 根据txt文件来移动图像，生成yolo数据集

2）VisDrone转Ultralytics YOLO格式

参考：https://blog.csdn.net/u012305034/article/details/134655981

自己的python脚本

3.在ultralytics/cfg/datasets/文件夹下建立自定义数据集的yaml文件

值得注意的是yaml文件里面的names的形式可以为：

# class names

names: ['windmill', 'stadium', 'overpass', 'ship', 'Expressway-Service-area', 'trainstation', 'storagetank', 'airport', 'tenniscourt', 'chimney', 'groundtrackfield', 'vehicle', 'golffield', 'bridge', 'dam', 'basketballcourt', 'Expressway-toll-station', 'harbor', 'airplane', 'baseballfield']

4.修改模型参数：ultralytics/cfg/models/v8/yolov8.yaml

修改nc参数，与上述的class names个数一致

值得注意的是，yolov8.yaml定义了五种不同复杂度的模型，yolov8n.yaml为最小模型

5. 训练模型

可以写一个train.py脚本，如以下形式：

##这里指定了训练模型的类型，输入图像大小，训练批次和学习率

from ultralytics import YOLO

import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Load a model

model = YOLO("yolov8n.yaml") # build a new model from YAML

model = YOLO("yolov8n.pt") # load a pretrained model (recommended for training)

model = YOLO("yolov8n.yaml").load("yolov8n.pt") # build from YAML and transfer weights

# Train the model

#results = model.train(data="VisDrone_tang.yaml", epochs=100, imgsz=640)

results = model.train(data="VisDrone_tang.yaml", epochs=100, imgsz=640,batch=32,lr0=0.001)

6.运行测试

1）单张图像测试：

yolo predict model=yolov8n.pt source='https://ultralytics.com/images/bus.jpg'

2）多张图像批量测试：

参考：

（1）直接解决yolov8的预测方法

yolov8的predict使用方法，更改predict.py的输出结果，输出label的真实坐标，保存图片和txt文档，图片中没有异物生成空的txt文档

https://blog.csdn.net/m0_57847261/article/details/134319876?spm=1001.2101.3001.6650.2&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7ECtr-2-134319876-blog-133688916.235%5Ev43%5Epc_blog_bottom_relevance_base7&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7ECtr-2-134319876-blog-133688916.235%5Ev43%5Epc_blog_bottom_relevance_base7&utm_relevant_index=5

（2）自己写的

运行前需要将DIOR_classes的类型顺序要与之前保持一致

Python detect_tang_v5.py

输出格式：类别+置信度+坐标（xmin,ymin,xmax,ymax）

6）得出map

(1) Python detect_tang_v5.py 这里要注意，一定要设置置信度为0.01，具体看mAP

(2) map_tang.py :

超参数：这里一定要这样设置，否则计算map极其慢

parser.add_argument('-na', '--no-animation',default=True, help="no animation is shown.", action="store_true")

parser.add_argument('-np', '--no-plot',default=False, help="no plot is shown.", action="store_true")

7） python版本TensorRT部署

参考：

视频教程：YOLOv8目标检测实战：ONNX模型转换及TensorRT部署

https://edu.csdn.net/learn/39292/635091

参考：https://zhuanlan.zhihu.com/p/667220840

Tensorrt安装及使用（python版本）：https://blog.csdn.net/stivory/article/details/135847529

1）ultralytics官方做法

（1）适用于 YOLOv8 模型的 ONNX 导出，可以学习ONNX是什么

网址：https://docs.ultralytics.com/integrations/onnx/

（2）YOLOv8 模型的 TensorRT 导出

网址：https://docs.ultralytics.com/integrations/tensorrt/#why-should-i-use-ultralytics-yolov8-cli

***环境部署：

pip install tensorrt -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnx -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxsim -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxruntime -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pycuda -i https://pypi.tuna.tsinghua.edu.cn/simple

1）ultralytics官方做法

（1）适用于 YOLOv8 模型的 ONNX 导出，可以学习ONNX是什么

网址：https://docs.ultralytics.com/integrations/onnx/

（2）YOLOv8 模型的 TensorRT 导出

网址：https://docs.ultralytics.com/integrations/tensorrt/#why-should-i-use-ultralytics-yolov8-cli

2）正常做法：pt先转ONNX格式，然后ONNX转换成TensorRT engine

3）TensorRT部署前和部署后的推理时间测试脚本

from ultralytics import YOLO

import time

#from openvino.inference_engine import IENetwork

#from onnxruntime import InferenceSession

old_time = time.time()

# Load a model

# model = YOLO('yolov8n.pt') # load an official model

#model = YOLO('best_openvino_model') # load a custom model

#model = YOLO('best.pt')

model = YOLO('best.engine')

# Predict with the model

results = model('datasets/images/1.jpg', save=True) # predict on an image

current_time = time.time()

print("推理耗时" + str(current_time - old_time) + "s")

8)YOLOv8+SAHI

参考：

使用YOLOv8+SAHI增强小目标检测效果并计算评估指标

https://blog.csdn.net/Dora_blank/article/details/140236280

pip install opencv-python sahi tabulate podm tqdm argparse -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install object_detection_metrics -i https://pypi.tuna.tsinghua.edu.cn/simple

ModuleNotFoundError: No module named 'podm.metrics'

解决办法：https://pypi.org/project/object-detection-metrics/0.1/

https://github.com/rafaelpadilla/Object-Detection-Metrics

pip install object_detection_metrics -i https://pypi.tuna.tsinghua.edu.cn/simple

9)YOLOv8+UI界面

参考：https://mbd.pub/o/bread/mbd-ZZyWlZ9y

########################################

requirements.txt 里面内容如下

certifi==2023.7.22

charset-normalizer==3.3.0

colorama==0.4.6

contourpy==1.1.1

cycler==0.12.1

fonttools==4.43.1

idna==3.4

importlib-resources==6.1.0

kiwisolver==1.4.5

matplotlib

###numpy==1.26.1

opencv-python==4.8.1.78

packaging==23.2

psutil==5.9.6

py-cpuinfo==9.0.0

pyparsing==3.1.1

python-dateutil==2.8.2

pyyaml==6.0.1

requests==2.31.0

scipy

seaborn==0.13.0

six==1.16.0

thop==0.1.1-2209072238

####torch==1.9.0

tqdm==4.66.1

typing-extensions==4.8.0

ultralytics==8.0.199

urllib3==2.0.6

zipp==3.17.0

PyQt5==5.15.2

pyqt5-tools==5.15.2.3.1

#################################

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n yolov8_ui python=3.8 -y

source activate yolov8_ui

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

pip uninstall opencv-python

pip install opencv-python-headless -i https://pypi.tuna.tsinghua.edu.cn/simple

解决expected np.ndarray (got numpy.ndarray)报错

https://blog.csdn.net/m0_56182552/article/details/140854252

QObject::moveToThread: Current thread is not the object`s thread. Cannot move to target thread

https://blog.csdn.net/ustczhng2012/article/details/115137510

pip uninstall opencv-python

pip install opencv-python-headless -i https://pypi.tuna.tsinghua.edu.cn/simple

四-1：YOLOv8-OBB推理详解及部署实现

https://gitcode.csdn.net/66262bb39c80ea0d2270d7e0.html?dp_token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpZCI6MzE5NjYxNSwiZXhwIjoxNzI0MTQ3NTIxLCJpYXQiOjE3MjM1NDI3MjEsInVzZXJuYW1lIjoid3V4aW56aGlzaG91In0.2hsLzH_RMer4-Zpckp3tLbCG1mFfE2XcjhHW20zTzbI

四-2：YOLOv8-OBB推理详解及部署实现

四-3：ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate ultralytics-8.3.16

1. 模型环境配置

https://github.com/ultralytics/ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n ultralytics-8.3.16 python=3.8 -y

source activate ultralytics-8.3.16

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

pip install ultralytics -i https://pypi.tuna.tsinghua.edu.cn/simple

##########TensorRT###############

pip install tensorrt -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnx -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxslim -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxruntime -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxruntime-gpu -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pycuda -i https://pypi.tuna.tsinghua.edu.cn/simple

###################SAHI################

pip install opencv-python sahi tabulate podm tqdm argparse -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install object_detection_metrics -i https://pypi.tuna.tsinghua.edu.cn/simple

需要进一步更新：pip install ultralytics -i https://pypi.tuna.tsinghua.edu.cn/simple

conda search gdal

(2.2.4,3.0.2)

conda install gdal==3.0.2

五、mmrotate

https://mmrotate.readthedocs.io/en/latest/install.html

1.：基于MMRotate训练自定义数据集 做旋转目标检测 2022-3-30

https://blog.csdn.net/qq_43581224/article/details/123838415?spm=1001.2101.3001.6661.1&utm_medium=distribute.pc_relevant_t0.none-task-blog-2%7Edefault%7ECTRLIST%7ETopBlog-1.topblog&depth_1-utm_source=distribute.pc_relevant_t0.none-task-blog-2%7Edefault%7ECTRLIST%7ETopBlog-1.topblog&utm_relevant_index=1

2.MMRotate 详细使用教程

https://blog.csdn.net/Sylvia_Lan/article/details/124008837?spm=1001.2101.3001.6650.1&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7Edefault-1.pc_relevant_default&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7Edefault-1.pc_relevant_default&utm_relevant_index=2

3.【AI目标检测】MMROTATE踩坑记录

https://blog.csdn.net/jaz_y/article/details/123983506

1. 模型环境配置

https://github.com/ultralytics/ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n mmrotate python=3.8 -y

source activate mmrotate

#conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=12.1 -c pytorch -c nvidia

conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.7 -c pytorch -c nvidia

pip install -U openmim -i https://pypi.tuna.tsinghua.edu.cn/simple

#安装mmcv-full ，一定要安装和cuda版本、pytorch版本对应的版本，如下命令

##pip install mmcv-full=={MMCV_VERSION} -f https://download.openmmlab.com/mmcv/dist/{CUDA_VERSION}/{TORCH_VERSION}/index.html

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

##Install mmcv

https://mmcv.readthedocs.io/zh-cn/latest/get_started/installation.html

pip install mmdet -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements/build.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

##其中pip install -v -e .or python setup.py develop将会安装为开发模式，所有对源码的修改都将生效。（pip install -v -e .最后的.表示当前路径，不要漏掉！）

#pip install -v -e . -i https://pypi.tuna.tsinghua.edu.cn/simple

python setup.py develop

pip install pillow -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install matplotlib -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install scikit-image -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install timm -i https://pypi.tuna.tsinghua.edu.cn/simple

bash: /home/zkti/data_16T/miniconda3/envs/mmrotate/bin/pip: /home/zkti/0.Test_tang/miniconda3/envs/mmrotate/bin/python: bad interpreter: No such file or directory

https://blog.csdn.net/weixin_44162814/article/details/143332667

关于MMrotate踩坑记录（ValueError：need at least one array to cancatenate）

https://blog.csdn.net/weixin_40493382/article/details/123897405?spm=1001.2101.3001.6650.1&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7EBlogCommendFromBaidu%7ECtr-1-123897405-blog-127943250.235%5Ev43%5Epc_blog_bottom_relevance_base5&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7EBlogCommendFromBaidu%7ECtr-1-123897405-blog-127943250.235%5Ev43%5Epc_blog_bottom_relevance_base5&utm_relevant_index=2

ICCV 2023 | LSKNet【保姆级】训练自己的目标检测模型

https://blog.csdn.net/weixin_51670564/article/details/132590489?utm_medium=distribute.pc_relevant.none-task-blog-2~default~baidujs_baidulandingword~default-0-132590489-blog-127943250.235^v43^pc_blog_bottom_relevance_base5&spm=1001.2101.3001.4242.1&utm_relevant_index=3

报错：RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

https://blog.csdn.net/weixin_51670564/article/details/132605642

【mmlab】mmrotate如何计算模型参数量

https://blog.csdn.net/weixin_44037500/article/details/141194256

https://mmrotate.readthedocs.io/zh-cn/v0.3.0/useful_tools.html

FPS计算

https://mmrotate.readthedocs.io/zh-cn/v0.3.0/useful_tools.html

成功解决error:unrecognized argument: --local-rank=0

https://blog.csdn.net/qq_43826289/article/details/131564035

2.数据集处理

1）类似DOTA大图的数据集

（1）都得先转换成DOTA格式

这里得 熟悉DOTA数据集格式和相关工具

DOTA遥感数据集以及相关工具DOTA_devkit的整理(踩坑记录）

DOTA数据集大图标签格式

注意点：大图格式里面，带imagesource和gsd两项

DOTA数据集切图后，小图切片的格式为

大图切图的程序，可以用/home/zkti/0.Test_tang/test_tang/detection_OBB/1.DOTA_bigimg/0.mmrotate/mmrotate_DIOR/tools/data/dota/split/img_split.py函数

里面需要配置好/home/zkti/0.Test_tang/test_tang/detection_OBB/1.DOTA_bigimg/0.mmrotate/mmrotate_DIOR/tools/data/dota/split/split_configs/ss_train.json等json文件

之后执行：

python mmrotate-main/tools/data/dota/split/img_split.py --base_json mmrotate-main/tools/data/dota/split/split_configs/ss_train.json

2）类似DIOR-R的小图数据集

（1）DIOR-R转换成DOTA小图数据集格式

使用XML2txt_8_DOTA.py脚本进行转换

3.配置参数进行训练

1）修改mmrotate_DIOR/mmrotate/datasets/dota.py的配置参数

（1）修改CLASSES

（2）将.png格式换成自定义数据集的图像格式

2）在0.mmrotate_DIOR文件夹下，新建configs_DIOR_R文件夹

3）将configs文件夹里面的_base_和你想要的方法，复制到到configs_DIOR_R文件夹下

4）修改configs_DIOR_R/_base_/datasets/dotav1.py的配置参数

（1）修改自定义数据集的路径：data_root = '/home/zkti/data_4T/1.datasets/2.Optical/0.DIOR/DIOR_R_DOTA/'

（2）修改img_scale=(800, 800)参数，将参数修改成与检测模型匹配的参数

（3）修改data = dict(

samples_per_gpu=2,

workers_per_gpu=2,

train=dict(

type=dataset_type,

ann_file=data_root + 'train/annfiles/',

img_prefix=data_root + 'train/images/',

pipeline=train_pipeline),

val=dict(

type=dataset_type,

ann_file=data_root + 'val/annfiles/',

img_prefix=data_root + 'val/images/',

pipeline=test_pipeline),

test=dict(

type=dataset_type,

ann_file=data_root + 'test/images/',

img_prefix=data_root + 'test/images/',

pipeline=test_pipeline))

5）修改configs_DIOR_R/oriented_rcnn/oriented_rcnn_r50_fpn_1x_dota_le90.py的配置参数

（1）修改num_classes=20，等于数据集类别个数，不加背景

（2）修改img_scale=(800, 800)参数，将参数修改成与检测模型匹配的参数

6)在0.mmrotate_DIOR文件下新建train_oriented_rcnn.sh的训练配置参数

CUDA_VISIBLE_DEVICES=0 python tools/train.py /home/zkti/0.Test_tang/test_tang/detection_OBB/0.DIOR_R/0.mmrotate_DIOR/configs_DIOR_R/oriented_rcnn/oriented_rcnn_r50_fpn_1x_dota_le90.py

4.常见问题：

1）ImportError: /home/zkti/0.Test_tang/miniconda3/envs/mmrotate/lib/python3.8/site-packages/mmcv/_ext.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZNK2at6Tensor10transpose_Ell

解决办法：https://github.com/open-mmlab/mmdetection/issues/4291

Same here, refer to the document: https://mmcv.readthedocs.io/zh-cn/latest/get_started/installation.html

reinstall mmcv with proper version according to your torch and cuda version, and the problem solved.

Note: In MMCV-v2.x, mmcv-full is rename to mmcv.

2）ImportError: cannot import name 'Config' from 'mmcv' (/home/zkti/0.Test_tang/miniconda3/envs/mmrotate/lib/python3.8/site-packages/mmcv/__init__.py)

解决办法：https://blog.csdn.net/M_TDM/article/details/133046534

python -c 'import torch;print(torch.__version__);print(torch.version.cuda)'

#安装mmcv-full ，一定要安装和cuda版本、pytorch版本对应的版本，如下命令

##pip install mmcv-full=={MMCV_VERSION} -f https://download.openmmlab.com/mmcv/dist/{CUDA_VERSION}/{TORCH_VERSION}/index.html

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

##Install mmcv

https://mmcv.readthedocs.io/zh-cn/latest/get_started/installation.html

3）KeyError: "DOTADataset: 'expressway-toll-station'"

解决方法：

mmrotate_DIOR/mmrotate/datasets/dota.py的CLASSES的大写都改成小写

再执行pip install -v -e . -i https://pypi.tuna.tsinghua.edu.cn/simple

六、HI-Diff-Hierarchical Integration Diffusion Model for Realistic Image Deblurring

# Clone the github repo and go to the default directory 'HI-Diff'.

git clone https://github.com/zhengchen1999/HI-Diff.git

cd HI-Diff

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n hi_diff python=3.9

conda create -n hi_diff_2 python=3.9 （可行）

source activate hi_diff

source activate hi_diff_2

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple #先把torch和torchvision注释，让其独立安装

#ERROR: No matching distribution found for tb-nightly

python -m pip install tb-nightly -i https://mirrors.aliyun.com/pypi/simple

在安装basicsr==1.4.2时候，发现安装始终出错。

https://blog.csdn.net/qq_45100200/article/details/141284517?spm=1001.2101.3001.6650.3&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7EYuanLiJiHua%7ECtr-3-141284517-blog-133925899.235%5Ev43%5Epc_blog_bottom_relevance_base7&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7EYuanLiJiHua%7ECtr-3-141284517-blog-133925899.235%5Ev43%5Epc_blog_bottom_relevance_base7&utm_relevant_index=4

pip install basicsr -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install安装的时候在Preparing metadata (setup.py) ...卡住

https://blog.csdn.net/sinat_29957455/article/details/130285223

pip install --upgrade pip setuptools wheel -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install --verbose basicsr --use-pep517 -i https://pypi.tuna.tsinghua.edu.cn/simple

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

# CUDA 11.3

conda install pytorch==1.9.0 torchvision==0.10.0 torchaudio==0.9.0 cudatoolkit=11.3 -c pytorch -c conda-forge（可行）

Generate image patches from GoPro dataset for training.

python generate_patches_gopro.py ##注释 from data.hdfs import copy, mkdir

#You can also specify --nproc_per_node=X with X>1.

解决报错:train.py: error: unrecognized arguments: --local-rank=1 或 local-rank=0

解决方法：https://blog.csdn.net/qq_43826289/article/details/132953452

python -m torch.distributed.run --nproc_per_node=2 --master_port=4321 train.py -opt options/train/GoPro_S1.yml --launcher pytorch

python -m torch.distributed.launch --nproc_per_node=1 --master_port=4321 train.py -opt options/train/GoPro_S2.yml --launcher pytorch

#python train.py -opt options/train/GoPro_S2.yml --launcher pytorch

pip install einops -i https://pypi.tuna.tsinghua.edu.cn/simple

A module that was compiled using NumPy 1.x cannot be run in

NumPy 2.0.2 as it may crash. To support both 1.x and 2.x

versions of NumPy, modules must be compiled with NumPy 2.0.

Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.

pip install numpy==1.26.4 -i https://pypi.tuna.tsinghua.edu.cn/simple

七、AutofocusSAR

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n AutofocusSAR python=3.8

source activate AutofocusSAR

pip install cython i

pip install torchbox torchsar -i https://pypi.tuna.tsinghua.edu.cn/simple

八、LoFormer

https://github.com/INVOKERer/LoFormer

https://github.com/INVOKERer/LoFormer

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate hi_diff_2

九。DEA-Net

DEA-Net: Single image dehazing based on detail-enhanced convolution and content-guided attention

(IEEE TIP 2024)

https://github.com/cecret3350/DEA-Net

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n DEA-Net python=3.8

source activate DEA-Net

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install matplotlib -i https://pypi.tuna.tsinghua.edu.cn/simple

CUDA_VISIBLE_DEVICES=0 python train.py --epochs 300 --iters_per_epoch 5000 --finer_eval_step 1400000 --w_loss_L1 1.0 --w_loss_CR 0.1 --start_lr 0.0001 --end_lr 0.000001 --exp_dir ../experiment/ --model_name DEA-Net-CR --dataset HAZE4K

File "/home/zkti/data_16T/2.Test_SAR_imaging/1.Deep_SAR_Imaging/DEA-Net-main/code/model/modules/cga.py", line 14, in forward

x2 = torch.concat([x_avg, x_max], dim=1)

AttributeError: module 'torch' has no attribute 'concat'

十、DehazeFormer

Vision Transformers for Single Image Dehazing

https://github.com/IDKiro/DehazeFormer?tab=readme-ov-file

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n DehazeFormer python=3.8

source activate DehazeFormer

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html

十一、C2PNet

https://github.com/YuZheng9/C2PNet

https://github.com/hpc203/C2PNet-onnxrun

[CVPR 2023] Curricular Contrastive Regularization for Physics-aware Single Image Dehazing

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n c2pnet python=3.7.11

source activate c2pnet

pip install torch==1.10.0+cu111 torchvision==0.11.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html

pip install -r requirements.txt

pip install tqdm -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install h5py -i https://pypi.tuna.tsinghua.edu.cn/simple

CUDA_VISIBLE_DEVICES=0 python main.py --crop --crop_size=240 --blocks=19 --gps=3 --bs=2 --lr=0.0001 --trainset='its_train' --testset='its_test' --steps=1000000 --eval_step=5000 --clcrloss --clip

十二、NAFNet

https://github.com/megvii-research/NAFNet/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n NAFNet python=3.9.5

source activate NAFNet

# CUDA 11.3

pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113

十二、mmediting

1. 模型环境配置

https://github.com/Libyte/mmediting/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n mmediting python=3.8 -y

source activate mmediting

conda install pytorch==2.0.0 torchvision==0.15.0 pytorch-cuda=11.8 -c pytorch -c nvidia

pip install -U openmim -i https://pypi.tuna.tsinghua.edu.cn/simple

#安装mmcv-full ，一定要安装和cuda版本、pytorch版本对应的版本，如下命令

##pip install mmcv-full=={MMCV_VERSION} -f https://download.openmmlab.com/mmcv/dist/{CUDA_VERSION}/{TORCH_VERSION}/index.html

pip install mmcv==2.0.0 -f https://download.openmmlab.com/mmcv/dist/cu118/torch2.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

##Install mmcv

https://mmcv.readthedocs.io/zh-cn/latest/get_started/installation.html

pip3 install -e . -i https://pypi.tuna.tsinghua.edu.cn/simple

AssertionError: mmcv==2.2.0 is used but incompatible. Please install mmcv-full>=[2, 0, -1, 1], <[2, 1, 0].

【Python报错】已解决 ModuleNotFoundError: No module named ‘transformers‘

pip install transformers -i https://pypi.tuna.tsinghua.edu.cn/simple

# single-gpu train

CUDA_VISIBLE=0 python tools/train.py configs/nafnet/nafnet_c64eb11128mb1db1111_8xb8-lr1e-3-400k_gopro.py --work-dir work_dirs/nafnet_c64eb11128mb1db1111_8xb8-lr1e-3-400k_gopro

CUDA_VISIBLE=0 python tools/train.py config_tang/nafnet/nafnet_SAR_Imaging.py

CUDA_VISIBLE=0 python tools/train.py config_tang/nafnet_v2/nafnet_SAR_Imaging.py

CUDA_VISIBLE=0 python tools/train.py config_tang/deblurganv2/deblurganv2_fpn-inception_1xb1_gopro.py

CUDA_VISIBLE=0 python tools/train.py configs/example_config.py --work-dir work_dirs/example

十三、Sparse_SAR_Imaging（师弟）

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n Sparse_SAR_Imaging python=3.11 -y

source activate Sparse_SAR_Imaging

conda install pytorch==2.2.2 torchvision==0.17.2 pytorch-cuda=12.1 -c pytorch -c nvidia

十四、SRCNN

https://github.com/yjn870/SRCNN-pytorch?tab=readme-ov-file

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate c2pnet

python train.py --train-file "BLAH_BLAH/91-image_x2_converted.h5" \

--eval-file "BLAH_BLAH/91-image_x2_converted.h5" \

--outputs-dir "BLAH_BLAH/outputs" \

--scale 2 \

--lr 1e-4 \

--batch-size 16 \

--num-epochs 400 \

--num-workers 8 \

--seed 123

python test.py --weights-file "/home/zkti/data_16T/2.Test_SAR_imaging/1.Deep_SAR_Imaging/3.image_super/0.SRCNN/code2/SRCNN-pytorch-master/BLAH_BLAH/outputs/x2/epoch_26.pth" \

--image-file "data/butterfly_GT.bmp" \

--scale 2

十五、SimDeblur

https://github.com/ljzycmd/SimDeblur

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate c2pnet （torch==1.10.0）

bash Install.sh -i https://pypi.tuna.tsinghua.edu.cn/simple

CUDA_VISIBLE_DEVICES=0 bash ./tools/train.sh ./configs/mscnn/mscnn_gopro.yaml 1

If you installed CuPy via wheels (cupy-cudaXXX or cupy-rocm-X-X), make sure that the package matches with the version of CUDA or ROCm installed.

pip install cupy-cuda122 -i https://pypi.tuna.tsinghua.edu.cn/simple

conda install cupy -i https://pypi.tuna.tsinghua.edu.cn/simple

十六、MIRNetv2

https://github.com/swz30/MIRNetv2/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate NAFNet

pip install matplotlib scikit-learn scikit-image opencv-python yacs joblib natsort h5py tqdm -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install einops gdown addict future lmdb numpy pyyaml requests scipy tb-nightly yapf lpips -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install tb-nightly -i https://mirrors.aliyun.com/pypi/simple

python setup.py develop --no_cuda_ext

https://github.com/Kobaayyy/Awesome-ECCV2024-ECCV2020-Low-Level-Vision/blob/master/ECCV2024.md

十七：sentinel

https://github.com/Rich-Hall/sentinel1decoder?tab=readme-ov-file

https://github.com/Rich-Hall/sentinel1Level0DecodingDemo

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate sentinel

conda create -n sentinel python=3.8

source activate sentinel

pip install nbconvert -i https://pypi.tuna.tsinghua.edu.cn/simple

jupyter nbconvert --to script sentinel1Level0DecodingDemo.ipynb

pip install object_detection_metrics -i https://pypi.tuna.tsinghua.edu.cn/simple

#pip install pillow matplotlib scikit-image opencv-python pandas -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pillow -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install matplotlib -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install scikit-image -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install tqdm -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install pandas -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install scipy -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install h5py -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install sympy -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install cupy -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install rasterio -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install tifffile -i https://pypi.tuna.tsinghua.edu.cn/simple

conda search gdal

conda install gdal==3.6.2

conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia

圣保罗

ASF下载，时间选取2020至2023，选取S3条带模式,HH,VV,Dual H,Dual V,Dual HV,Dual VH

https://github.com/jmfriedt/sentinel1_level0?tab=readme-ov-file#point-like-target-the-sao-paulo-stripmap-sm-dataset

十七-1 SSFocus

https://github.com/sirbastiano/SSFocus/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate activate SSFocus

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n SSFocus python=3.8

pip install pillow matplotlib scikit-image opencv-python tqdm pandas scipy h5py sympy cupy netcdf4 h5netcdf rasterio rioxarray rioxarray scikit-learn xarray geopandas asf_search seaborn tzlocal regex -i https://pypi.tuna.tsinghua.edu.cn/simple

conda search gdal

conda install gdal==3.6.2

运行具有“SSFocus (Python 3.8.20)”的单元格需要ipykernel包。

运行以下命令，将 "ipykernel" 安装到 Python 环境中。

命令: "conda install -n SSFocus ipykernel --update-deps --force-reinstall"

git https://github.com/Rich-Hall/sentinel1decoder

cd sentinel1decoder

pip install .

pip install torch==2.0.0+cu118 torchvision==0.15.1+cu118 -f https://download.pytorch.org/whl/torch_stable.html -i https://pypi.tuna.tsinghua.edu.cn/simple

十八：CI-CDNet

https://github.com/bianlab/CI-CDNet/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate c2pnet

十九：ADMM-CSNet

https://github.com/yangyan92/Pytorch_ADMM-CSNet/tree/main

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate hi_diff

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

二十：mmdetection

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate mmdetection

https://blog.csdn.net/xiao_9626/article/details/136803906

首先确定已安装好MMDetection环境，能正常运行 demo/image_demo.py 输出内容。

这里需要注意一点，如果高版本MMDetection安装输出有问题，建议安装低版本的库 mmcv<2.0以及mmdet<3.0,以及更低版本的pytorch<2.0，这一点非常关键，在MMDetection高版本上运行不了的程序，在低版本上是可以运行的。

————————————————

版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。

原文链接：https://blog.csdn.net/xiao_9626/article/details/136803906

1. 模型环境配置

https://github.com/ultralytics/ultralytics

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create -n mmdet_214 python=3.8 -y

source activate mmdet_214

#conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=12.1 -c pytorch -c nvidia

#conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.7 -c pytorch -c nvidia

conda install pytorch==1.13.0 torchvision==0.14.0 torchaudio==0.13.0 pytorch-cuda=11.7 -c pytorch -c nvidia

import torch

print(torch.cuda.is_available())

print(torch.cuda.get_device_name())

print(torch.cuda.current_device())

print(torch.cuda.device_count())

print(torch.__version__)

print(torch.version.cuda)

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu117/torch1.13.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install mmdet -i https://pypi.tuna.tsinghua.edu.cn/simple

#安装mmcv-full ，一定要安装和cuda版本、pytorch版本对应的版本，如下命令

##pip install mmcv-full=={MMCV_VERSION} -f https://download.openmmlab.com/mmcv/dist/{CUDA_VERSION}/{TORCH_VERSION}/index.html

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements/build.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

python setup.py develop

##Install mmcv

https://mmcv.readthedocs.io/zh-cn/latest/get_started/installation.html

pip install mmcv==2.2.0 -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html -i https://pypi.tuna.tsinghua.edu.cn/simple

问题：AssertionError: MMCV==2.2.0 is used but incompatible. Please install mmcv>=2.0.0rc4, <2.2.0.

解决方法： pip uninstall mmcv

pip install mmcv==2.0.0 -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html

pip install mmdet<3.0 -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements/build.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

##其中pip install -v -e .or python setup.py develop将会安装为开发模式，所有对源码的修改都将生效。（pip install -v -e .最后的.表示当前路径，不要漏掉！）

#pip install -v -e . -i https://pypi.tuna.tsinghua.edu.cn/simple

python setup.py develop

pip install pillow -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install matplotlib -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install scikit-image -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install timm

1.1.3 训练自己数据集

Faster RCNN训练pascalvoc数据集:

https://zhuanlan.zhihu.com/p/162730118

【mmdetection】使用自定义的coco格式数据集进行训练及测试

https://blog.csdn.net/gaoyi135/article/details/90613895

https://zhuanlan.zhihu.com/p/101983661

https://blog.csdn.net/qq122716072/article/details/115245410

SourceURL:file:///home/zkti/data_16T/0.Test_tang/test_tang/经典目标检测模型使用教程-20240715.docx

Faster RCNN训练pascalvoc数据集:

https://zhuanlan.zhihu.com/p/162730118

【mmdetection】使用自定义的coco格式数据集进行训练及测试

https://blog.csdn.net/gaoyi135/article/details/90613895

https://zhuanlan.zhihu.com/p/101983661

https://blog.csdn.net/qq122716072/article/details/115245410

1.1.3.1 不重写Dataset加载模块

1．训练pascalvoc数据集

参考https://blog.csdn.net/weixin_41010198/article/details/106258366

mmdetectionv2.14版本

1)新建文件夹

在mmdetection内新建configs_tang_voc_COCO和work_dirs两个文件夹

configs_tang_voc_COCO：放入读入数据库和运行pascalvoc数据集的模型

work_dirs:放入训练时的缓存模型

2）复制文件夹

首先将mmdetection/config中的‘_base_’和‘pascal_voc’ 文件夹复制到configs_tang_voc_COCO内

‘_base_/datasets’：存放读取voc数据集的程序voc0712.py

‘pascal_voc’ :存放运行pascalvoc数据集的模型

3)修改python程序

mmdetection/configs_tang_voc_COCO/_base_/datasets/voc0712.py:data_root='文件夹位置'

mmdetection/mmdet/core/evaluation/class_names.py: 修改类别def voc_classes()

mmdetection/mmdet/datasets/voc.py : 修改类别CLASSES=()

模型修改

mmdetection/configs_tang_voc_COCO/pascal_voc/retinanet_r50_fpn_1x_voc0712.py: num_classes=20

4)运行训练

python tools/train.py configs_tang_voc_COCO/pascal_voc/retinanet_r50_fpn_1x_voc0712.py --gpus 1

mmdetection中出现 loss为 nan的情况

参考网址：http://events.jianshu.io/p/f111eaf4393d

5)运行测试

Python demo_batch_test_mmdet.py

(1) demo_batch_test_mmdet.py 这里要注意，一定要设置置信度为0.01，具体看mAP

(2) map_tang.py :

超参数：这里一定要这样设置，否则计算map极其慢

图像，真值，输出结果路径

7）DIOR的test数据集测试结果（训练集为train,不是trainval）

模型:

2. 训练coco数据集

1)coco数据集格式

coco

-anancontions

-train2017

-val2017

-test2017

2)pascalvoc转coco格式

(1)创建coco格式文件夹

(2)查看数据集类别： voc_labelname.py

(3)移动/VOCdevkit/VOC2007/JPEGImages到

coco/train2017,coco/val2017,coco/test2017文件夹：select_test.py

(4)pascalvoc转coco格式程序：2.voc2coco_mmdetection.py

（参考mmdetection/tools/dataset_converters/pascal_voc.py)

（5）COCO格式可视化：4.visual_coco.py

3)新建文件夹

在mmdetection内新建configs_tang_voc_COCO和work_dirs两个文件夹

configs_tang_voc_COCO：放入读入数据库和运行coco数据集的模型

work_dirs:放入训练时的缓存模型

4）复制文件夹

首先将mmdetection/config中的‘_base_’和‘retinanet’ 文件夹复制到configs_tang_voc_COCO内

‘_base_/datasets’：存放读取voc数据集的程序voc0712.py

‘retinanet’ :存放运行coco数据集的retinanet模型

5)修改python程序

mmdetection/configs_tang_voc_COCO/_base_/datasets/coco_detection.py:data_root='文件夹位置'

mmdetection/configs_tang_voc_COCO/_base_/datasets/coco_instance.py:data_root='文件夹位置'

#mmdetection/mmdet/core/evaluation/class_names.py: 修改类别def coco_classes()

mmdetection-main/mmdet/evaluation/functional/class_names.py: 修改类别def coco_classes()

mmdetection/mmdet/datasets/coco.py : 修改类别CLASSES=()

注意：只要修改一次，就得执行python setup.py install命令，否则mmdete会按照以前的类别计算，报错。

模型修改

mmdetection/configs_tang_voc_COCO/_base_/models/retinanet_r50_fpn.py: num_classes=80

6)运行训练

configs_tang_voc_COCO/retinanet/retinanet_r50_caffe_fpn_1x_coco.py

python tools/train.py configs_tang_voc_COCO/retinanet/retinanet_r50_caffe_fpn_1x_coco.py --gpus 1

7)运行测试

Python demo_batch_test_mmdet.py

(1) demo_batch_test_mmdet.py 这里要注意，一定要设置置信度为0.01，具体看mAP

(2) map_tang.py :

超参数：这里一定要这样设置，否则计算map极其慢

图像，真值，输出结果路径

mmdetection ValueError: need at least one array to concatenate 已解决

https://blog.csdn.net/weixin_52869984/article/details/137930813

21：X-AnyLabeling

https://github.com/CVHub520/X-AnyLabeling/blob/main/docs/zh_cn/get_started.md

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

source activate x-anylabeling

python anylabeling/app.py

export PATH=/home/zkti/data_16T/miniconda3/bin:$PATH

conda create --name x-anylabeling python=3.9 -y

source activate x-anylabeling

pip install onnxruntime -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install onnxruntime-gpu -i https://pypi.tuna.tsinghua.edu.cn/simple

pip install -r requirements-gpu.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

pyrcc5 -o anylabeling/resources/resources.py anylabeling/resources/resources.qrc

python anylabeling/app.py