Skip to content

JeremyKTH/Plastic-Bale-Image-Recognition

Repository files navigation

HK - Volvo CE Software

Table of contents

Plastic_Bale_Recognition

Introduction

In order to recognize plastic bale, a model of yolov3 is used. The model is trained on a custom made dataset and recognizes one class, plastic bales. When running yolo, the software creates a bounding box surrounding the plastic bale.

Datasets

The datasets that is being used is created by the team. It consists of images taken from the internet of the plastic bale as well as images of our own plastic bale model. In order to get more data, we have augumented the images with 90 degree flips to get more images.

Results

Introduction_Mapping

A 2D map is created through SLAM. The camera being used is a ZED2, which has an IMU. Therefore, it is translated into laserscan in order to use gmapping.

Navigation

Requirements

  • Python 3.5+
  • PyTorch 3.8.1
  • Ubuntu 18.04 or 20.04
  • ROS Melodic or Neotic
  • CUDA >= 10.2
  • cuDNN >= 8.0.2
  • OpenCV >= 2.4
  • CMake >= 3.18

Contacts

About

A collaboration project with Volvo CE developing an automated guided vehilce

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors