Skip to content

Nx1021/RKDT

Repository files navigation

RGB-Based Set Prediction Transformer of 6D Pose Estimation for Robotic Grasping Application

Paper url: TODO

Framework

framework

Requirements

  • h5py==3.7.0
  • matplotlib==3.5.2
  • MyLib==0.0.1
  • numpy==1.22.3
  • open3d==0.17.0
  • opencv_python==4.6.0.66
  • pandas==1.4.3
  • Pillow==9.1.1
  • Pillow==10.0.0
  • python_Levenshtein==0.21.1
  • scikit_learn==1.2.2
  • scipy==1.8.1
  • sko==0.5.7
  • tensorboard==2.10.0
  • torch==1.12.1+cu113
  • torchvision==0.13.1+cu113
  • tqdm==4.65.0
  • ultralytics==8.0.119

Installation

This code has been tested on :

  • Ubuntu22.04 with Cuda 12.5, Python3.9 and Pytorch 1.12.1.
  • Win10 with Cuda 12.4, Python3.9 and Pytorch 1.12.1.

Get the code

git clone https://github.com/Nx1021/RKDT.git

Create new Conda environment.

conda create -n rkdt python=3.9
conda activate RKDT

Install torch

conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.6 -c pytorch -c conda-forge

Others

pip install -r requirements.txt

Usage

Checkpoint

TODO

Dataset

TODO

Train

TODO

Test

TODO

Results

Object ADD(-S) 2D-proj
ape 91.05 99.81
bench. 97.19 96.61
cam 88.39 95.67
can 98.23 99.61
cat 88.22 99.90
driller 97.82 98.41
duck 80.09 99.81
eggbox 99.25 99.06
glue 96.62 99.71
holep. 79.35 99.52
iron 98.47 98.26
lamp 98.46 97.79
phone 94.79 99.15
average 92.92 98.72

Citation

TODO

About

Object Landmarks Detection Transformer

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages