https://github.com/dev-michael-schmidt/eyeball-sam

A home grown real-time object dectecion for embedded devices using YOLOv7
https://github.com/dev-michael-schmidt/eyeball-sam

Last synced: 3 months ago
JSON representation

A home grown real-time object dectecion for embedded devices using YOLOv7

• Host: GitHub
• URL: https://github.com/dev-michael-schmidt/eyeball-sam
• Owner: dev-michael-schmidt
• License: gpl-3.0
• Created: 2023-01-26T02:43:59.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-07-02T15:32:58.000Z (12 months ago)
• Last Synced: 2025-01-14T11:16:23.516Z (5 months ago)
• Language: Jupyter Notebook
• Size: 36.6 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# eyeball-sam

![](https://github.com/MichaelSchmidt82/eyeball-sam/blob/main/content/cool_demo.gif)

**What:**  Real-time object detection, person detection, and face recognition using YOLOv7 in TensorFlow Lite targeted for devices at the edge with Google Coral hardware.

## Requirements:

### Software

- 🖥️ Ubuntu 20.04

- 🐍️ Python 3.8

- 📦️ See requirements.txt, there are a lot.

- 📷️ It is recommend to ~[build OpenCV from source](https://docs.opencv.org/4.x/d7/d9f/tutorial_linux_install.html) for local testing (or just in general).~ use the requirements.txt version.

### Hardware

- 🌊️ [Google Coral](https://coral.ai/). They have low-wattage USB and M.2 TPUs. A must for real-time video processing.

Usage:

1. Create a virtual environment and `pip install -r requirements.txt`.

2. Run the `create_tf_lite.ipynb` notebook to download use the model weights. This notebook will convert ONNX format to tf-lite.

3. Run `tfl_yolov7_main.py`.

Note: by default, openCV will use your wedcam (`cv2.VideoCapture(0)`)

This project was updated on 01/29/2024