# YOLO V7 [TOC] --- ## 訓練帳號 :::success ccit.edu.tw ::: # 考慮google colab ## YOLOV7環境安裝 [YOLOv7 on Windows (一) 建立環境與執行](https://www.larrysprognotes.com/YOLOv7_1/) 1. 透過`nvidia-smi`,檢查自己顯卡版本(必須要是獨立顯卡NVIDIA<=可以透過工作管理員查看) 2. 至[NVIDIA CUDA Toolkit Release Notes](https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html)查看自己的顯卡版本對應的CUDA套件版本,確認後再到[INSTALLING PREVIOUS VERSIONS OF PYTORCH](https://pytorch.org/get-started/previous-versions/)尋找對應的Pytorch版本進行下載 + 我自己對應版本 `pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2` 3. 檢查是否安裝成功 :::info >>> import torch >>> torch.cuda.is_available() True >>> torch.cuda.get_device_name() 'NVIDIA GeForce MX330' >>>  ::: --- ## 用YOLOV7訓練自己想要的資料 ### 1. 訓練資料標註 1. ` pip install simple_image_download==0.4`,可以透過這個套件去google image抓取圖片,keywords可以換成自己想要的詞 ```python= from simple_image_download import simple_image_download as simp response = simp.simple_image_download keywords=["speed limit sign"] for kw in keywords: response().download(kw, 300) my_downloader = simp.simple_image_download() my_downloader.directory = "grab_image" ``` + 註:https://github.com/RiddlerQ/simple_image_download 2. `pip install labelimg`,透過labelimg標記圖片,檔名要改成yolo 3. 圖庫 https://universe.roboflow.com/ ### 2. 訓練 [YOLOv7 on Windows (二) 訓練自己的資料集](https://www.larrysprognotes.com/YOLOv7_2/) 1. 看顯卡內存 2. #### yolo訓練時報錯問題 1. AssertionError: Invalid device id => 修改方式 --device 1 :::spoiler  ::: 2. ## 參考資料 + [【小白教学】如何用YOLOv7训练自己的数据集](https://zhuanlan.zhihu.com/p/547878330) + [Official YOLO v7 Custom Object Detection Tutorial | Windows & Linux](https://www.youtube.com/watch?v=-QWxJ0j9EY8) ## function ConnectButton(){ console.log("Connect puushed"); document.querySelector("#top-toolbar > ccolab-connect-button").shadowRoot.querySelector("#connect").click() } setInterval(ConnectButton,60000);
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up