# Tensorflow Lite 實作 Object detection ###### tags: `object detection` ## 待辦 * 用神經加速棒來跑 mobile net or YOLO ## 遠端控制 RaspberryPi ```python '''登入''' ssh pi@raspberrypi.local #password:raspberrypi '''檔案傳輸''' scp [來源端檔案路徑] pi@[172.20.10.3]:/home/pi/Documents ``` (參考影片) https://www.youtube.com/watch?v=aimSGOAUI8Y * 樹莓派[基本設定](https://medium.com/ljlstyle/%E6%A8%B9%E8%8E%93pi-%E5%88%9D%E5%A7%8B%E7%92%B0%E5%A2%83%E8%A8%AD%E5%AE%9A-b2df781f9108) * Raspberry Pi 開發環境:[see](https://docs.donkeycar.com/guide/install_software/#install-donkeycar-on-windows) ## keras-yolo3 * 在 mac 上執行,辨識一張照片需要 0.6 ~ 1 秒 * keras-yolo3 無法在 Raspberry Pi 3 上面跑,會killed ## 在樹莓派上實現 object detection * 主要參考網站:[HERE](https://www.rs-online.com/designspark/raspberry-pi-4intel-cn) * 參考網站:[link](https://chtseng.wordpress.com/2017/09/15/在樹莓派上運行object-detection-api/) * 樹莓派上安裝 tensorflow [link](https://github.com/samjabrahams/tensorflow-on-raspberry-pi) * https://github.com/EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi/blob/master/Raspberry_Pi_Guide.md ## 安裝 Tensorflow Lite * 教學:[在Raspberry Pi4上安裝Tensorflow Lite for python以及Coral USB Accelerator套件](https://blog.cavedu.com/2019/10/09/raspberry-pi4-tensorflow-lite-for-python-coral-usb-accelerator/) * Terminal 執行 * sudo apt-get update * sudo apt-get upgrade -y * sudo apt-get install build-essential * 根據python的版本,到官網下載ARM32的那個安裝包 * 在樹莓派執行 sudo pip3 install 安裝包名稱 * 看到tflite-runtime就over了! ```python '''如果沒有辦法 upgrade 的話,可以刪除這些檔案''' cd /var/lib/dpkg/updates rm -r ./* ``` * 使用範例:[HERE](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/examples/python) * https://github.com/tensorflow/examples/blob/master/lite/examples/object_detection/raspberry_pi/README.md ## MobileNet https://github.com/Zehaos/MobileNet ## 如何使用神經網路加速棒 * 參考文章 * [邊緣運算】如何在樹莓派使用 NCS 神經運算棒](https://makerpro.cc/2019/01/how-to-use-ncs-with-raspberry-pi/) * https://chtseng.wordpress.com/2018/05/16/movidius-neural-stick的設定安裝及使用/ * [教學影片](https://www.youtube.com/watch?v=joElT3UfspA)
×
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