yolov5和yolov5-face环境搭建和常见踩坑 - plus studio

2024-02-28 21:18:59 浏览数 (2)

yolov5环境搭建

在随便哪新建一个requirements.txt文件 内容是

代码语言：text复制
# YOLOv5 requirements
# Usage: pip install -r requirements.txt

# Base ----------------------------------------
matplotlib>=3.2.2
numpy>=1.18.5
opencv-python>=4.1.1
Pillow>=7.1.2
PyYAML>=5.3.1
requests>=2.23.0
scipy>=1.4.1
torch>=1.7.0
torchvision>=0.8.1
tqdm>=4.64.0
protobuf<=3.20.1  # https://github.com/ultralytics/yolov5/issues/8012

# Logging -------------------------------------
tensorboard>=2.4.1
# wandb
# clearml

# Plotting ------------------------------------
pandas>=1.1.4
seaborn>=0.11.0

# Export --------------------------------------
# coremltools>=5.2  # CoreML export
# onnx>=1.9.0  # ONNX export
# onnx-simplifier>=0.4.1  # ONNX simplifier
# nvidia-pyindex  # TensorRT export
# nvidia-tensorrt  # TensorRT export
# scikit-learn==0.19.2  # CoreML quantization
# tensorflow>=2.4.1  # TFLite export (or tensorflow-cpu, tensorflow-aarch64)
# tensorflowjs>=3.9.0  # TF.js export
# openvino-dev  # OpenVINO export

# Extras --------------------------------------
ipython  # interactive notebook
psutil  # system utilization
thop>=0.1.1  # FLOPs computation
# albumentations>=1.0.3
# pycocotools>=2.0  # COCO mAP
# roboflow

然后在当前目录下打开命令行，创建一个环境

代码语言：text复制
conda create -n yolov5 python

创建好环境之后，激活环境

代码语言：text复制
conda activate yolov5

然后安装依赖

代码语言：text复制
pip install -r requirements.txt

安装完成后代码就可以运行了

划分数据集

新建一个split_train_val.py文件，内容如下

代码语言：text复制
import os
import shutil
import random

def split_dataset(src_folder, dest_folder, ratio):
    images_folder = os.path.join(src_folder, "images")
    labels_folder = os.path.join(src_folder, "labels")

    if not os.path.exists(images_folder) or not os.path.exists(labels_folder):
        raise Exception("Source folder doesn't exist.")
    if not os.path.exists(dest_folder):
        os.makedirs(dest_folder)

    train_folder = os.path.join(dest_folder, "train")
    val_folder = os.path.join(dest_folder, "val")
    if not os.path.exists(train_folder):
        os.makedirs(train_folder)
    if not os.path.exists(val_folder):
        os.makedirs(val_folder)

    train_images_folder = os.path.join(train_folder, "images")
    train_labels_folder = os.path.join(train_folder, "labels")
    val_images_folder = os.path.join(val_folder, "images")
    val_labels_folder = os.path.join(val_folder, "labels")
    if not os.path.exists(train_images_folder):
        os.makedirs(train_images_folder)
    if not os.path.exists(train_labels_folder):
        os.makedirs(train_labels_folder)
    if not os.path.exists(val_images_folder):
        os.makedirs(val_images_folder)
    if not os.path.exists(val_labels_folder):
        os.makedirs(val_labels_folder)

    images = [f for f in os.listdir(images_folder) if f.endswith(".bmp") ]
    num_images = len(images)

    for i, image in enumerate(images):
        image_path = os.path.join(images_folder, image)
        label_path = os.path.join(labels_folder, os.path.splitext(image)[0]   ".txt")
        if random.uniform(0, 1) < ratio:
            dest_images_folder = train_images_folder
            dest_labels_folder = train_labels_folder
        else:
            dest_images_folder = val_images_folder
            dest_labels_folder = val_labels_folder
        shutil.copy2(image_path, os.path.join(dest_images_folder, image))
        shutil.copy2(label_path, os.path.join(dest_labels_folder, os.path.splitext(image)[0]   ".txt"))
        print("Copied {}/{} images".format(i   1, num_images))

if __name__ == "__main__":
    src_folder = ""# 原始数据集的路径
    dest_folder = ""# 分割后的数据集的路径
    ratio = 0.8  # 将 80% 的图片分到训练集，20% 的图片分到验证集

    split_dataset(src_folder, dest_folder, ratio)

划分训练集和验证集，运行split_train_val.py，传入刚才保存的文件夹路径，会将图片和标签划分到一个新的文件夹

代码语言：text复制
- data
    - train
        - images
        - labels
    - val
        - images
        - labels

yolov5常见踩坑

not enough values to unpack (expected 2, got 0)

我们需要检查一下我们标记的txt文件 举个例子

这是我们需要的标记格式

代码语言：text复制
0 0.5 0.5 0.5 0.5

这是错误的标注格式

代码语言：text复制
0 0.5 0.5 0.5 0.5

问题就出在了最后一行的n上，我们删除最后一行就可以了。我用chatGPT写了一个函数来做这件事

代码语言：text复制
# 去除txt文件中的空行
def remove_empty_lines(file_path):
    with open(file_path, 'r') as f:
        lines = f.readlines()
    with open(file_path, 'w') as f:
        for line in lines:
            if len(line)>3:
                f.write(line)

AssertionError: No results.txt files found in /content/yolov5-face/runs/train/exp, nothing to plot.

代码语言：text复制
Traceback (most recent call last):
File "train.py", line 513, in
train(hyp, opt, device, tb_writer, wandb)
File "train.py", line 400, in train
plot_results(save_dir=save_dir) # save as results.png
File "/content/yolov5-face/utils/plots.py", line 393, in plot_results
assert len(files), 'No results.txt files found in %s, nothing to plot.' % os.path.abspath(save_dir)
AssertionError: No results.txt files found in /content/yolov5-face/runs/train/exp, nothing to plot.

出现这个问题的原因是此代码块未运行

代码语言：text复制
# Results
       if ckpt.get('training_results') is not None:
           with open(results_file, 'w') as file:
               file.write(ckpt['training_results'])  # write results.txt

如果你只使用单 GPU 并设置 epoch <20，这个块将不起作用。解决方案是设置epoch>20。

gitpython找不到对应版本

代码语言：text复制
ERROR: Could not find a version that satisfies the requirement gitpython>=3.1.30 (from versions: 0.1.7, 0.2.0b1, 0.3.0b1, 0.3.0b2, 0.3.1b2, 0.3.2rc1, 0.3.2, 0.3.2.1, 0.3.3, 0.3.4, 0.3.5, 0.3.6, 0.3.7, 1.0.0, 1.0.1, 1.0.2, 2.0.0, 2.0.1, 2.0.2, 2.0.3, 2.0.4, 2.0.5, 2.0.6, 2.0.7, 2.0.8, 2.0.9.dev0, 2.0.9.dev1, 2.0.9, 2.1.0, 2.1.1, 2.1.3, 2.1.4, 2.1.5, 2.1.6, 2.1.7, 2.1.8, 2.1.9, 2.1.10, 2.1.11, 2.1.12, 2.1.13, 2.1.14, 2.1.15, 3.0.0, 3.0.1, 3.0.2, 3.0.3, 3.0.4, 3.0.5, 3.0.6, 3.0.7, 3.0.8, 3.0.9, 3.1.0, 3.1.1, 3.1.2, 3.1.3, 3.1.4, 3.1.5, 3.1.6, 3.1.7, 3.1.8, 3.1.9, 3.1.10, 3.1.11, 3.1.12, 3.1.13, 3.1.14, 3.1.15, 3.1.16, 3.1.17, 3.1.18, 3.1.19, 3.1.20)
ERROR: No matching distribution found for gitpython>=3.1.30

python版本过低，升级即可。

path txt 环境搭建 数据 join

0 人点赞