109 lines
2.4 KiB
Markdown
109 lines
2.4 KiB
Markdown
|
# YoloV8 與 YoloV11 安裝教學
|
||
|
|
|
||
|
|
## 基礎安裝
|
||
|
|
|
||
|
|
1. 先找到nvdia控制面板
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/1.png" >
|
||
|
|
|
||
|
|
|
||
|
|
2. 系統資訊=> 點元素 (本人顯卡1070 版本CUDA 11.1.102)
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/2.png" >
|
||
|
|
|
||
|
|
|
||
|
|
3. 官網尋找自己的版本,不能下載超過自己的版本 (https://developer.nvidia.com/cuda-toolkit-archive) (本人下載版本11.1.1)
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/3.png" >
|
||
|
|
|
||
|
|
|
||
|
|
4. 下載CUDNN,選擇自己CUDA的版本可以的 (https://developer.nvidia.com/rdp/cudnn-archive) (本人下載版本8.9.6.50 CUDA 11 )
|
||
|
|
|
||
|
|
5. 將CUDNN 解壓縮複製 三個資料 bin,include,lib
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/4.png" >
|
||
|
|
|
||
|
|
|
||
|
|
6. 將CUDNN 複製的檔放到CUDA 只接貼上取代就好 (本人CUDA位置 C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1)
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/5.png" >
|
||
|
|
|
||
|
|
|
||
|
|
7. 安裝好後 打開cmd 確認安裝完畢 nvcc -V
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/6.png" >
|
||
|
|
|
||
|
|
## 安裝Python 以及 Anaconda 和 Pycharm
|
||
|
|
1. Python 官網(https://www.python.org/downloads/)
|
||
|
|
2. Anaconda 官網(https://www.anaconda.com/download)
|
||
|
|
3. Pycharm 官網 (https://www.jetbrains.com/pycharm)
|
||
|
|
4. Anaconda與 Pycharm 都要先註冊好
|
||
|
|
|
||
|
|
## Pycharm 與YoloV8 與 YoloV11 以及安裝Pytorch
|
||
|
|
### 因YoloV8 跟 YoloV11 前半段安裝都是相同的,只有安裝模型時的不同
|
||
|
|
1. 利用Anaconda 方式選擇Python 3.10版
|
||
|
|
2. Pycharm 的Cmd Terminal 前面要顯示你剛剛命名的虛擬機 本人是(yoloV8) (這步驟很重要,用好才可安裝Pytorch)
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/7.png" >
|
||
|
|
|
||
|
|
|
||
|
|
3. 接著安裝Pytorch 官網(https://pytorch.org/)
|
||
|
|
4. 選擇好版本複製,記得pip3去掉3 留pip (本人 pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118)
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/8.png" >
|
||
|
|
|
||
|
|
|
||
|
|
5. 將複製好的貼上Pycharm安裝
|
||
|
|
6. 安裝好打pip list 顯示
|
||
|
|
|
||
|
|
|
||
|
|
<img src="images/9.png" >
|
||
|
|
|
||
|
|
|
||
|
|
7. 確定安裝好後 接著安裝
|
||
|
|
```
|
||
|
|
pip install ultralytics
|
||
|
|
|
||
|
|
```
|
||
|
|
8. 測試並安裝模型 (下載n模型或s模型都可以,本人都用s模型)
|
||
|
|
|
||
|
|
YoloV8
|
||
|
|
```
|
||
|
|
yolo predict model=yolov8n.pt source='https://ultralytics.com/images/bus.jpg'
|
||
|
|
|
||
|
|
```
|
||
|
|
YoloV11
|
||
|
|
```
|
||
|
|
yolo predict model=yolo11n.pt source='https://ultralytics.com/images/bus.jpg'py
|
||
|
|
|
||
|
|
```
|
||
|
|
```
|
||
|
|
yolo predict model=yolo11s.pt source='https://ultralytics.com/images/bus.jpg'py
|
||
|
|
|
||
|
|
```
|
||
|
|
9. 安裝標記
|
||
|
|
```
|
||
|
|
pip install labelimg
|
||
|
|
|
||
|
|
```
|
||
|
|
10. 執行
|
||
|
|
```
|
||
|
|
labelimg
|
||
|
|
|
||
|
|
```
|
||
|
|
## 講解模型
|
||
|
|
|
||
|
|
主要以 n, s, m, l, 和 x 表示
|
||
|
|
|
||
|
|
<img src="images/10.png" >
|
||
|
|
|
||
|
|
基本上都用n 跟 s 就好 (本人都用s,準度很高,顯卡(1070)輕鬆跑)
|