NVIDIA Jetson Nano學習筆記（七）：即時影像偵測暨辨識系統(PiCamera+OpenCV+TensorFlow Lite+Firebase)
本文章假設讀者已完成TF Runtime的安裝，並且已經自行裝上Pi Camera、設定好Firebase的RealTime Database及Storage。
（１）尚未安裝TF Runtime的讀者可以參考：
NVIDIA Jetson Nano學習筆記（四）：安裝與執行Tensorflow Lite Model官方範例
（２）尚未設定好Firebase的RealTime Database及Storage的讀者可以參考：
Learning Firebase(1)：Create Your First Project
Learning Firebase(2)：CRUD Our Database with Python
Learning Firebase(3)：Upload Image to Firebase with Python
在NVIDIA Jetson Nano學習筆記（五）：即時影像分類系統(PiCamera+OpenCV+TensorFlow Lite+Firebase)一文當中，我和大家介紹了如何在Jetson Nano中，透過PiCamera進行即時的Image Classification。
在NVIDIA Jetson Nano學習筆記（六）：即時影像偵測系統(PiCamera+OpenCV+TensorFlow Lite)一文當中，我和大家介紹了如何在Jetson Nano中，透過PiCamera進行即時的Object Detection。
正文
本文延續前兩次的主題，結合了Object Detection和Image Classification。
Google官方提供的Object Detection Pre-trained Model準確度還不錯，未來解決我們Self-trained Model誤判的情形，我們先用物件偵測來判斷是否為指定的Label，如果是，我們再透過影像分類，來做詳細Label Name分類。
GitHub Repo：
程式功能：
[1] Load Pre-trained (Object Detection) and Self-trained (Image Classification)TFLite Model with Argument.[2] Read image from PiCamera with OpenCV to do Real-Time Object Detection.[3] If detect specific object ("bird" in the code), save the image.[4] Use Self-trained Model to do Image Classification on the image with OpenCV.[5] Upload the Image and classification result (LabelName, ScoreValue, Time, Pubic-Access Image Url) to Firebase Database[6] Save the above result (LabelName, ScoreValue, Time, Pubic-Access Image Url) as a csv file with append mode.[7] Once the image and data have been uploaded to Firebase, delete the local images to prevent running out of disk space.
使用方法：
python3 object_detection_and_image_classification.py專案結構：
Folder:
Sample_TFLite_model/:Contain the object detection model and label
File:
object_detection_and_image_classification.py:Our main program of this project.TFLite_Read_Image.py:Read Image with OpenCV to Image Classification.test.tflite:Image Classification TFLite Model.test.txt:Image Classification TFLite label.firebase_key.json:If you want to use firebase to store your data, you should have it.
程式碼：