/coral_live_stream_demo

Object detection live streaming using Coral USB Accelerator

Primary LanguagePythonMIT LicenseMIT

Coral object classification model live stream

Description

Stream USB camera with Coral Edge TPU based object detection/classification over network. The project is dependent on Google's Coral USB Accelerator.

Credits

Base code took from log0's video_streaming_with_flask_example. Coral object classification originally from Google LLC.

Usage

  1. Install the Edge TPU runtime and Python library following https://coral.withgoogle.com/docs/accelerator/get-started/
  2. Install Python dependencies: cv2, flask.
  3. Download pretrained Edge TPU sample models using: sh download_models.sh
  4. Run "python3 main.py", or run "sh ./start.sh" in the background.
  5. Navigate the browser to the http://ip-adress:5000.

Switch models

  1. Switch between detection/classification by modifying main.py imports
  2. Change the model used for detection/classification by modifying the config.py

Reference

  1. https://github.com/log0/video_streaming_with_flask_example
  2. https://coral.withgoogle.com/docs/edgetpu/api-intro