M-Calib: A Monocular 3D Object Localization using 2D Estimates for Industrial Robot Vision System

Implementation code for our paper "M-Calib: A Monocular 3D Object Localization using 2D Estimates for Industrial Robot Vision System"

Thanh Nguyen Canh, Du Trinh Ngoc, Xiem HoangVan, "Monocular 3D Object Localization using 2D Estimates for Industrial Robot Vision System," Journal of Automation, Mobile Robotics and Intelligent Systems, 2024. [Journal of Automation, Mobile Robotics and Intelligent Systems] [Citation]

Citation

@article{Canh2024,
  title = {Monocular 3D Object Localization using 2D Estimates for Industrial Robot Vision System},
  ISSN = {},
  url = {},
  DOI = {},
  journal = {Journal of Automation, Mobile Robotics and Intelligent Systems},
  publisher = {Industrial Research Institute for Automation and Measurements PIAP, Poland},
  author = {Thanh, Nguyen Canh and Du, Trinh Ngoc and Xiem HoangVan},
  year = {2024},
  month = jul,
  pages = {}
}

Our proposed:

Requirements

• python 3.7
• torch 1.7.1
• tensorboard

M-Calib Datasets

There are two different datasets collected by the authors

The related datasets can be found at:

• 1. Object Detection dataset: (https://app.roboflow.com/uet-jvl1l/m-calib/1).
• 2. Object Segmentation dataset: (https://app.roboflow.com/uet-jvl1l/mcalibsegment/1).

Usage: M-Calib (the inference)

• 1. Step 1: Clone this repo

git clone https://github.com/thanhnguyencanh/MonoCalibNet
cd MonoCalibNet

• 2. Step 2: Creating a model

Option 1:

• Use a pre-trained model: model

• Modify path in cfg.py

Option 2:

• Training new model using Object_Detection and Instance_Segmentation

• Convert to Onnx model

• 3. Step 3: Testing

mkdir model
mkdir dataset

• Set pat in cfg.py

python3 run_chessboard_yolov5.py