rs-opencv

Introduction

Using:

• Ubuntu 20.04 (Requirement) I think, it is supported in Ubuntu 18.04 has to be test
• Python 3.9 (Requirement)
• TensorRT
• Pytorch
• OpenCV
• RealSense Camera (D435)

⚡ The script is working. It need to improve in the part of handle the recognized objects. Also the recognition model has to be improve.

⚡ Support for regular cameras has to be included.

Requirements

• Ubuntu 20.04 or Ubuntu 18.04
• Python3.9
• CUDA cores (NVIDIA GPU or Jetson Nano)

Test if CUDA is available

python3.9
>>import torch
>>torch.cuda.is_available()
True

If result is False:

Check: After Installing Ubuntu 20.04

Setup

git clone https://github.com/john-wick1999/rs-opencv.git
cd rs-opencv/
virtualenv -p /bin/python3.9 venv
source venv/bin/activate
pip install -r requirements.txt

Start

Activate virtual environment

source venv/bin/activate

Add models

Models are located in Data folder with name (new model change best.engine):

• best.engine
• coco.yaml

Run script

Activate venv.

Connect camera.

python detect.py