We provide our dataset IGD for 6D object pose estimation that focuses on metallic objects. It introduces complex scenarios with different lighting conditions and backgrounds. The dataset follows the BOP format, making the data also available for 2D object detection and segmentation.

Scene.zip

contains the Blender file, where all the objects are present in one file
used to generate the dataset with BlenderProc (https://github.com/DLR-RM/BlenderProc)

Models.zip

Contains all the models in .ply format for training and evaluation of the results
Contains the models_info.json, which has more information about the model's geometry

ustom_Dataset.zip

This is the main training dataset, which contains 90k images for six different scenes

Custom_Dataset_Material.zip

This is a reduced dataset, where every folder beside the RGB image from the scenes contains a folder no_material where the same scene is present but with adjusted material parameters.

Test.zip

Contains the test images with their annotations. This is the primary test set!
Contains test images for all six scenes

Test_Tray.zip

It contains test images with a reflective surface below every scene to create more reflections.
Can be useful for evaluating different scenarios
Contains test images for all six scenes

Test_NR.zip

Contains test images where no reflections are present within the scenes.
It can be useful for evaluating different scenarios
Contains test images for all six scenes

If you find our data helpful, please consider citing our work

      title={Improving 6D Object Pose Estimation of metallic Household and Industry Objects}, 
      author={Thomas Pöllabauer and Michael Gasser and Tristan Wirth and Sarah Berkei and Volker Knauthe and Arjan Kuijper},
      year={2025},
      eprint={2503.03655},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.03655}, 
}

Credits:

all the objects used within this dataset are obtained from sketchfab.com.
The objects are evenly split into three categories: Can, Household, and Industry.

Objects

Cans

"Redbulls" (https://skfb.ly/oxHzu) by antoineallardchoquette

"Monster Energy Concept" (https://skfb.ly/oUPtC) by zxckurayami

"Monster Energy Pipeline Punch 355ml JPN" (https://skfb.ly/6QWWT) by VRC-IW

"energy banks" (https://skfb.ly/ovS8B) by sleepy owl

"Pepsi Can" (https://skfb.ly/6CCON) by Yanez Designs

"Coca cola cans" (https://skfb.ly/oQAYP) by Safat_Rules

"Pepsi Cans" (https://skfb.ly/oQrty) by Safat_Rules

"Guinness Beer Can" (https://skfb.ly/6A78J) by nickmisketis

"CC0 - Tin Can 4" (https://skfb.ly/oEUJQ) by plaggy

Household

"Scissors" (https://skfb.ly/oD7YI) by sweedboy69

"Camping silverware pack" (https://skfb.ly/o8RIq) by Guy in a Poncho

"Soup Ladle" (https://skfb.ly/oFIMo) by jones.troyjr1221

"Golden tablespoon, dessert spoon and teaspoon" (https://skfb.ly/osTQP) by Nortenko Dmytro

"Metal cutlery" (https://skfb.ly/opVvZ) by Okayest

"Kitchen Pack" (https://skfb.ly/6DGWq) by GRIP420

"Thermos Bottle" (https://skfb.ly/6XTL9) by Loopenkoopen

"Mug" (https://skfb.ly/6SVNu) by Kobayashi Danaki

"Sketchfab Texturing Challenge: Sports Car Base" (https://skfb.ly/6uppt) by Renafox

"Canarian Cafe - Milk Jug" (https://skfb.ly/opAtH) by Lanzaman

"T-tool" (https://skfb.ly/6WQPL) by achchu.zats

Industry

"Nuts And Bolts" (https://skfb.ly/orAOv) by Ryan King Art

"Imbus Bolt - Imbus Civata" (https://skfb.ly/6G6zQ) by wuhle

"NEMA17 Plastic Bracket Support L Type" (https://skfb.ly/oD7Zq) by YouniqueIdeaStudio

"Spindle Bracket Holder" (https://skfb.ly/oErP6) by YouniqueIdeaStudio

"Basic Kit-Bash Vol 2" (https://skfb.ly/6FUUW) by Somersby

"Basic Kit-Bash Vol 3" (https://skfb.ly/6FVus) by Somersby

"Barometer Tray" (https://skfb.ly/6CDnA) by Aaron Rovinsky

"XY Joiner" (https://skfb.ly/6VtJS) by matt_nolan11

"Hard Surface Combining Shapes" (https://skfb.ly/6wo6E) by 3DTrip

"Hard Surface KitBash #1" (https://skfb.ly/onxPs) by Dr. Props

"Gears" (https://skfb.ly/6WrAC) by metalheadqx

"Flange. Car parts" (https://skfb.ly/6WVWF) by trofitec

Surface

"SIMPLE tray" (https://skfb.ly/oSpsp) by max.zolot

Datasets:

tpoellabauer
/

IGD

Contents