graph LR;
GitHub_Actions[GitHub Actions Workflow] --> Cron[Cron Job Run Every 15 Minutes];
Cron --> Python_Script[Python Script];
Python_Script --> NYC_Traffic_Cam_Footage[Real-Time NYC Traffic Cam Footage];
NYC_Traffic_Cam_Footage --> Object_Detection[Object Detection using PyTorch Vision];
Object_Detection --> Detection_Output[Detection Output];
Detection_Output -->Insert_DB[Insert into SQLite DB];
This workflow is quite linear and is effectively automated ETL in the sense that objects are being detected in NYC traffic cams every 15 minutes and the output is being stored in a database (utc_timestamp
, camera_location
, detected_object
, score
, and bounding_box
); no other image data is retained or stored, as demonstrated in the Colab notebook provided below:
Note: I let this experiment run for a couple weeks and have since archived the project, meaning that I am no longer collecting and storing data here every 15 minutes.