🏁 添加在 win 上运行的脚本，并优化了脚本内容

.github/workflows/docs-pnpm.yml

@@ -13,10 +13,12 @@ jobs:
     steps:
       - name: Checkout this
         # @see https://github.com/actions/checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0 # 如果未启用 lastUpdated，则不需要
 
       - name: Install Node.js
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
         with:
           node-version: 20
 

.gitignore

@@ -6,6 +6,7 @@ tmps/
 temp/
 temps/
 resource/
+projects/
 
 
 # Vuepress/Vitepress

docs/.vitepress/config.mts

@@ -8,7 +8,17 @@ export default defineConfig({
   themeConfig: {
     // https://vitepress.dev/reference/default-theme-config
     nav: [
-      { text: 'Home', link: '/' },
+      { text: '首页', link: '/' },
+      { text: '项目源码', link: 'https://github.com/HenryZhuHR/deep-object-detect-track' },
+      {
+        text: '项目文档',
+        items: [
+          { text: '安装环境', link: '/install' },
+          { text: '数据集制作', link: '/dataset' },
+          { text: '模型训练', link: '/train' },
+          { text: '模型部署', link: '/deploy' },
+        ]
+      }
     ],
 
     sidebar: [
@@ -19,7 +29,6 @@ export default defineConfig({
           { text: '数据集制作', link: '/dataset' },
           { text: '模型训练', link: '/train' },
           { text: '模型部署', link: '/deploy' },
-          // { text: '目标检测YOLOv5项目', link: '/yolo' },
         ]
       }
     ],

docs/dataset.md

@@ -15,7 +15,7 @@ outline: deep
 DATASET_DIR=/path/to/dataset
 ```
 
-> 需要注意的是，数据集通常需要放置在项目外的路径，例如 `~/data` （推荐）。如果放置在项目内，导致编辑器对于项目的索引过大，会导致编辑器卡顿
+> 需要注意的是，数据集通常需要放置在项目外的路径，例如 `~/data` 或 `$HOME/data` （推荐）（win 下为 `$env:USERPROFILE/data`）。如果放置在项目内，导致编辑器对于项目的索引过大，会导致编辑器卡顿
 
 这里准备好了一个示例数据集，可以下载
 

docs/deploy.md

@@ -13,6 +13,13 @@ outline: deep
 
 <!--@include: ./download-pretrian.md-->
 
+同时，也提供了一些转化好的模型（从 ultralytics/yolov5(v7.0) 的 [yolov5](https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s) 导出），可以直接使用进行部署
+
+```shell
+bash scripts/download-premodels.sh
+#zsh scripts/download-premodels.sh # zsh
+```
+
 ## 导出模型
 
 提供一个导出脚本 `scripts/train.sh`，复制一份到项目目录下进行自定义修改（推荐）
@@ -285,17 +292,22 @@ python3 export.py \
 ```
 - `--device` 指定 GPU 设备，有多少张卡就写多少张卡，用逗号分隔
 
-修改 `infer.py` 文件，指定模型路径
+修改 `infer.py` 文件
 ```python
 ## -- TensorRT
-args.model=".cache/yolov5/yolov5s.engine"
 detector = backends.TensorRTBackend()
 ```
 
+执行推理
+```shell
+python infer.py --model .cache/yolov5/yolov5s.engine
+```
+
 
 推理速度
 
 | Tensor 版本 | 模型    | 推理时间(ms) | 设备           |
 | ----------- | ------- | ------------ | -------------- |
 | 10.0(.1.6)  | yolov5s | 3.26 ~ 3.72  | RTX 4090 (24G) |
+| 10.0(.1.6)  | yolov5s | 10.41 ~ 10.7  | RTX 3050 (8G) |
 | 8.6(.1.6)   | yolov5s | 3.40 ~ 3.85  | RTX 4090 (24G) |

docs/install.md

@@ -31,23 +31,33 @@ cd deep-object-detect-track
 
 > 后续的脚本基于 `deep-object-detect-track` 目录下执行
 
-如果未能获取子模块，可以手动获取
-```shell
+如果未能获取子模块，可以手动获取，如果 `git submodule` 无法获取，可以使用 `git clone` 获取
+
+::: code-group
+
+```shell [git submodule]
 # in deep-object-detect-track directory
 git submodule init
 git submodule update
 ```
 
+```shell [git clone]
+git clone https://github.com/ultralytics/yolov5.git projects/yolov5
+```
+
+:::
+
+
 ## 系统要求
 
 ### 操作系统
 
 
-项目在 Linux(Ubuntu) 和 MacOS 系统并经过测试 ✅ ，经过测试的系统：
-- Ubuntu 22.04 jammy (CPU & GPU)
-- MacOS (CPU)
+项目在 Linux(Ubuntu) 和 MacOS 系统并经过测试 ，经过测试的系统：
+- ✅ Ubuntu 22.04 jammy (CPU & GPU)
+- ✅ MacOS (CPU)
 
-Windows 系统暂未测试也无适配计划，如果需要在 Windows 系统上运行，可以使用 WSL2 或者根据提供的脚本手动执行，不接受 Windows 系统的问题反馈
+> 项目不支持 Windows 系统 ❌ ，如果需要在 Windows 系统上运行，可以使用 WSL2 或者根据提供的脚本手动执行；虽然已经测试通过，但是不保证所有功能都能正常运行，因此不接受 Windows 系统的问题反馈
 
 ### GPU
 
@@ -63,6 +73,8 @@ export PATH="$CUDA_HOME/bin:$PATH"
 export LD_LIBRARY_PATH="$CUDA_HOME/lib64:$LD_LIBRARY_PATH"
 ```
 
+> 事实上，Pytorch 1.8 开始就会在安装的时候自动安装对应的 CUDA Toolkit，因此不需要手动安装 CUDA Toolkit，因此可以跳过这一步
+
 MacOS 系统不支持 CUDA Toolkit，可以使用 CPU 训练模型 (Yolov5 项目暂不支持 MPS 训练)，但是推理过程可以使用 Metal ，参考 [*Introducing Accelerated PyTorch Training on Mac*](https://pytorch.org/blog/introducing-accelerated-pytorch-training-on-mac/#getting-started) 和 [*MPS backend*](https://pytorch.org/docs/stable/notes/mps.html#mps-backend)
 
 
@@ -191,3 +203,6 @@ pip install -r requirements/requirements.train.txt
 - `onnx`: ONNX 格式的模型导出支持任意设备，需要取消注释，并且其他导出依赖于 ONNX 模型
 - `coremltools`: 必须依赖于 MacOS 系统
 - `nvidia-*`: 确保硬件支持 NVIDIA GPU
+
+
+

infer.py

@@ -2,6 +2,7 @@
 import os
 from typing import Dict
 import cv2
+import tqdm
 import yaml
 
 from dlinfer.detector import DetectorInferBackends
@@ -26,25 +27,21 @@ def main() -> int:
     ## ------ ONNX ------
     onnx_backend = backends.ONNXBackend
     print("-- Available devices:", providers := onnx_backend.SUPPORTED_DEVICES)
-    detector = onnx_backend(
-        device=providers, inputs=["images"], outputs=["output0"]
-    )
+    detector = onnx_backend(device=providers, inputs=["images"], outputs=["output0"])
 
     ## ------ OpenVINO ------
     # ov_backend = backends.OpenVINOBackend
     # print("-- Available devices:", ov_backend.query_device())
-    # detector = ov_backend(device="AUTO")
+    # detector = ov_backend(device="GPU.0")
 
     ## ------ TensorRT ------
     # detector = backends.TensorRTBackend()
     # =======================================================
 
     detector.load_model(args.model, verbose=True)
 
-    with open(".cache/yolov5/yolov5s_openvino_model/yolov5s.yaml", "r") as f:
-        label_map: Dict[int, str] = yaml.load(f, Loader=yaml.FullLoader)[
-            "names"
-        ]
+    with open(".cache/yolov5/coco.yaml", "r") as f:
+        label_map: Dict[int, str] = yaml.load(f, Loader=yaml.FullLoader)["names"]
         label_list = list(label_map.values())
         # print(label_list)
 
@@ -57,23 +54,26 @@ def main() -> int:
 
     # -- do inference
     print("-- do inference")
-    for i in range(10):
+    pbar = tqdm.tqdm(range(100))
+    total_sum_time = 0
+    _cnt = 0
+    for i in pbar:
         start_time = cv2.getTickCount()
+        # -- preprocess
         input_t, scale_h, scale_w = Process.preprocess(img)  # B C H W
+        # -- inference
         output_t = detector.infer(input_t)
-
         end_time = cv2.getTickCount()
         infer_time = (end_time - start_time) / cv2.getTickFrequency() * 1000
-
+        # -- postprocess
         preds = Process.postprocess(output_t)
-
         end_time = cv2.getTickCount()
         total_time = (end_time - start_time) / cv2.getTickFrequency() * 1000
-
-        print(
-            f"Time infer/total: {infer_time:.2f}/{total_time:.2f} ms",
-        )
+        total_sum_time += total_time
+        _cnt += 1
+        pbar.set_description(f"Time infer/total: {infer_time:.2f}/{total_time:.2f} ms")
     # -- mark
+    print(f"-- Average time: {total_sum_time / _cnt:.2f} ms")
     Process.mark(img, preds, label_list, scale_h, scale_w)
     cv2.imwrite(save_path := "tmp/out.jpg", img)
     print(f"-- output saved to '{save_path}'")

scripts/base.ps1

@@ -0,0 +1,38 @@
+# =============== Color Print ===============
+$DEFAULT = [char]27 + '[0m'
+$RED = [char]27 + '[00;31m'
+$GREEN = [char]27 + '[00;32m'
+$YELLOW = [char]27 + '[00;33m'
+$CYAN = [char]27 + '[00;36m'
+
+function print_base     { param($color, $type, $message) Write-Host -NoNewline "$color- [$type] $message"; Write-Host $DEFAULT }
+function print_info     { param($message) print_base $CYAN      "INFO"      $message }
+function print_tip      { param($message) print_base $YELLOW    "TIP"       $message }
+function print_success  { param($message) print_base $GREEN     "SUCCESS"   $message }
+function print_warning  { param($message) print_base $YELLOW    "WARNING"   $message }
+function print_error    { param($message) print_base $RED       "ERROR"     $message }
+
+function run_script {
+    param($scriptPath)
+    if (-not (Test-Path $scriptPath -PathType Leaf)) {
+        Write-Host "$scriptPath not found"
+        exit 1
+    }
+    . $scriptPath  # 指定 powershell 执行脚本
+}
+
+$variablesCustomPath = "scripts/variables.custom.ps1"
+$variablesDefaultPath = "scripts/variables.ps1"
+
+if (Test-Path $variablesCustomPath -PathType Leaf) {
+    run_script $variablesCustomPath
+    print_success "Loaded custom variables: '$variablesCustomPath'"
+}
+elseif (Test-Path $variablesDefaultPath -PathType Leaf) {
+    run_script $variablesDefaultPath
+    print_success "Loaded default variables: '$variablesDefaultPath'"
+}
+else {
+    Write-Host "'$variablesCustomPath' or '$variablesDefaultPath' not found"
+    exit 1
+}

scripts/create-python-env.ps1

@@ -0,0 +1,68 @@
+# 启动 PowerShell：打开 PowerShell 控制台。可以通过在Windows搜索框中键入“PowerShell”并选择相应的选项来启动。
+
+# 设置执行策略（可选）：如果您的系统的执行策略不允许运行脚本，您可能需要先更改执行策略。可以通过在管理员权限下打开的 PowerShell 中运行以下命令来更改执行策略：
+
+# powershell
+# 复制代码
+# Set-ExecutionPolicy RemoteSigned
+# 这将允许运行本地的、来自 Internet 的已签名脚本，但不会执行未签名的本地脚本。
+
+# Check if scripts\base.ps1 exists
+if (-not (Test-Path "scripts\base.ps1")) {
+    Write-Host "scripts\base.ps1 not found"
+    exit 1
+}
+
+. "scripts\base.ps1"
+
+# Check if ENV_PATH is set
+if (-not $script:ENV_PATH) {
+    print_error "ENV_PATH is not set. Please set it in 'scripts\variables.ps1'"
+    exit 1
+}
+
+$script:ENV_PATH = "$script:ENV_PATH.conda"
+
+# Check if ENV_PATH directory exists
+if (-not (Test-Path $script:ENV_PATH -PathType Container)) {
+    $DEFAULT_PYTHON_VERSION = 3.10
+    if (-not $script:CUSTOM_PYTHON_VERSION) {
+        $ENV_PYTHON_VERSION = $DEFAULT_PYTHON_VERSION
+    } else {
+        $ENV_PYTHON_VERSION = $script:CUSTOM_PYTHON_VERSION
+    }
+    conda create -p $script:ENV_PATH -y python=$ENV_PYTHON_VERSION
+    print_success "Create Python environment in '$script:ENV_PATH'"
+} else {
+    print_info "Conda environment '$script:ENV_PATH' already exists."
+}
+
+conda activate $script:ENV_PATH
+
+# 该安装命令参考 https://pytorch.org/get-started/locally/
+pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121
+pip install -r projects/yolov5/requirements.txt
+pip install -r requirements/requirements.train.txt
+
+conda deactivate
+
+function print_activate_env_message {
+    Write-Host ""
+    print_info "Run command below to activate the environment:"
+    Write-Host ""
+    Write-Host "source ~/.$(Split-Path -Leaf $PROFILE)"
+    Write-Host "conda activate $script:ENV_PATH"
+    Write-Host ""
+    print_info "Then run command below to deactivate the environment:"
+    Write-Host ""
+    Write-Host "source ~/.$(Split-Path -Leaf $PROFILE)"
+    Write-Host "conda deactivate"
+    Write-Host ""
+}
+print_activate_env_message
+
+# cd .\project\deep-object-detect-track\
+# conda activate .\.env\deep-object-detect-track.conda\
+# python .\export.py --weights "C:\Users\29650\project\deep-object-detect-track\.cache\yolov5\yolov5s.pt" --data data/coco128.yaml --simplify --include onnx openvino engine --device 0
+# python .\export.py --weights "C:\Users\29650\project\deep-object-detect-track\.cache\yolov5\yolov5s.pt" --data data/coco128.yaml --include onnx openvino engine --device 0
+# python .\export.py --weights "C:\Users\29650\project\deep-object-detect-track\.cache\yolov5\yolov5s.pt" --data data/coco128.yaml --include engine --device 0

scripts/create-python-env.sh

@@ -30,7 +30,8 @@ case $key in
     echo "Usage: create-python-env.sh [OPTIONS]"
     echo "Options:"
     echo "  -e, --env-provider <provider>  Set the environment provider (venv or conda)"
-    echo "  -ni, --no-install              Skip Python requirements installation"
+    echo "  -i, --install                  Install Python requirements"
+    echo "  -v, --verbose                  Enable verbose mode"
     echo "  -h, --help                     Show this help message and exit"
     exit 0
     ;;

scripts/download-premodels.sh

@@ -0,0 +1,22 @@
+if [ ! -f "scripts/base.sh" ]; then
+    echo "scripts/base.sh not found"
+    exit 1
+fi
+source scripts/base.sh
+
+tag_name=v1.0.0
+base_url=https://github.com/HenryZhuHR/deep-object-detect-track/releases/download/${tag_name}
+
+
+weights_dir=$CACHE_DIR/yolov5
+[ ! -d ${weights_dir} ] && mkdir -p ${weights_dir}
+
+wget -c ${base_url}/coco.yaml -P ${weights_dir}
+wget -c ${base_url}/yolov5s.onnx -P ${weights_dir}
+
+
+ov_dir=${weights_dir}/yolov5s_openvino_model
+[ ! -d ${ov_dir} ] && mkdir -p ${ov_dir}
+
+wget -c ${base_url}/yolov5s.bin -P ${ov_dir}
+wget -c ${base_url}/yolov5s.xml -P ${ov_dir}

scripts/download-yolov5.sh

@@ -15,7 +15,7 @@ all_model_list=(
 )
 
 weights_dir=$CACHE_DIR/yolov5
-mkdir -p ${weights_dir}
+[ ! -d ${weights_dir} ] && mkdir -p ${weights_dir}
 
 model_list=($1)
 

scripts/train.sh

@@ -49,3 +49,5 @@ python3 train.py \
     --device $TRAIN_DEVICE \
     --workers 8 \
     --project $PROJECT_HOME/tmp/runs/train
+
+# python train.py --data $env:USERPROFILE/data/bottle-organized/dataset.yaml  --cfg models/yolov5s.yaml --weights "../../.cache/yolov5/yolov5s.pt" --epochs 10 --batch 4 --device 0