Manuel

[Manuel @StoiberMa]

The M3T tracker estimates the pose and configuration of objects in 3D space and supports a wide range of kinematic structures, object characteristics, and camera setups. Its efficient and modular architecture enables numerous applications in robotics and AR.

[Manuel @StoiberMa]

Happy to announce that the Robot Tracking Benchmark (RTB) and our multi-body, multi-modality, and multi-camera 3D object tracker M3T are now open source.

Paper: https://t.co/cPdKBnHXxN
Code: https://t.co/XfMJzo5L7T
Dataset: https://t.co/qa5Z6DUd6X
Video: https://t.co/AUmDlep7Ih

[Manuel @StoiberMa]

The Robot Tracking Benchmark features highly-realistic sequences with multiple robots, difficulty levels, and image qualities. It was created using BlenderProc. The dataset allows to extensively measure, compare, and ablate the performance of algorithms for multi-body objects.

[Manuel @StoiberMa]

@Kirumang Thanks! We used an Azure Kinect for the video.

[Manuel @StoiberMa]

Happy to share our work "Iterative Corresponding Geometry (ICG)" accepted to #CVPR2022! SOTA 6DoF tracking, requiring only the object geometry and 1.3ms per frame on a single CPU core.

arXiv: https://t.co/l17JpfZxCH
Code: https://t.co/VJXMxh5sB3
Video: https://t.co/Kv8kfIxzfS

[Manuel @StoiberMa]

Finally, 𝐒𝐑𝐓𝟑𝐃: 𝐀 𝐒𝐩𝐚𝐫𝐬𝐞 𝐑𝐞𝐠𝐢𝐨𝐧-𝐁𝐚𝐬𝐞𝐝 𝟑𝐃 𝐎𝐛𝐣𝐞𝐜𝐭 𝐓𝐫𝐚𝐜𝐤𝐢𝐧𝐠 𝐀𝐩𝐩𝐫𝐨𝐚𝐜𝐡 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐑𝐞𝐚𝐥 𝐖𝐨𝐫𝐥𝐝, which is the follow-up work to our award-winning @accv2020 paper, is available on GitHub and arXiv: https://t.co/zhdk5Ie4Id

[Maximilian Denninger @DenningerMax]

Times flies by. BlenderProc is out in version 1.8.0 and this is a big one. We now have a complete web page for our documentation.

We also added a random room constructor and added the haven websites. For more see the change log.

https://t.co/47gle0bndb

https://t.co/dLTV362Mc7

[ACCV2020 @accv2020]

#ACCV2020 Best Paper (Saburo Tsuji Award)

Stoiber et al., “A Sparse Gaussian Approach to Region-Based 6DoF Object Tracking”
https://t.co/laycwTHQdz

[Manuel @StoiberMa]

For more details, please view the talk and visit our Live Session @accv2020. Also, check out the following resources:
- real-world video: https://t.co/h6b0MEGDPE
- simulation video: https://t.co/U1ArhVCC4h
- code: https://t.co/OMkS9PS5Yo
- paper: https://t.co/XyiBzj7kp7

[Manuel @StoiberMa]

We are thankful to #ACCV2020 chairs and reviewers for being awarded the Best Paper Award @accv2020. Our work on 6DoF object tracking shows that by using a sparse formulation with a Gaussian likelihood function and Newton optimization, region-based methods become super efficient!

[Antonin Raffin @araffin2]

From our lab: a fast and robust 6DoF Object Tracking that requires only a monocular RGB camera and the 3D object model (to appear in ACCV 2020).

Video (sim): https://t.co/UJrDQr6NBL
Video (real): https://t.co/l6MGktI5z8
Code: https://t.co/f9zGRPdQjc