Urs

We lanuched the project page for 3D-MuPPET, a framework to estimate and track 3D poses of up to 10 pigeons at interactive speed. We show that 3D-MuPPET also works in natural environments without model fine-tuning on additional annotations.

Check it out. Especially our Hugging Face Demo!

Check out our latest pre-print on arXiv https://sciencecast.org/casts/8g1z2dov39ms

We present 3D-MuPPET, a framework to estimate and track 3D poses of up to 10 pigeons at interactive speed. We show that 3D-MuPPET also works in natural environments without model fine-tuning on additional annotations.

This year’s CVPR took place in Vancouver, Canada from June 18-22.

We presented three of our works accepted in the Workshop „CV4Animals“ held on June 18. We had two posters in the AM session and one in the PM session.

I really enjoyed the venue, the poster sessions, the discussions and the people that I met.

We finally provide our pre-trained weights for pigeons to use with I-MuPPET.

https://zenodo.org/record/7037589#.ZHcsFl7P0uU

Also, our pigeon tracking benchmark is available now.

zenodo.org/record/7038391#.ZHcs8V7P0uU

Our results, that you can improve upon, can be found in our paper.

And last but not least, our single pigeon data with RGB images and corresponding keypoint annotations is available now.

zenodo.org/record/7989831#.ZHcwO17P0uU

You can check it out with I-MuPPET.

We hope that this will help to boost the study of collective behaviour.

Three of our papers were accepted at the CVPR workshop „CV4Animals: Computer Vision for Animal Behavior Tracking and Modeling“ that will take place on June 18, 2023 in Vancouver, Canada.

3D-MuPPET: 3D Multi-Pigeon Pose Estimation and Tracking

I-MuPPET: Interactive Multi-Pigeon Pose Estimation and Tracking
https://urs-waldmann.github.io/i-muppet/

Improving Unsupervised Label Propagation for Pose Tracking and Video Object Segmentation
https://urs-waldmann.github.io/improving-unsupervised-label-propagation/

There were several reports about our last work „Neural Puppeteer“.

The University of Konstanz published a press release that you can find here. There is also a German version here.

The Centre for the Advanced Study of Collective Behaviour reported in their news as well about NePu.

And the Innovation Origins further coveraged our work: https://innovationorigins.com/en/selected/ai-captures-animal-motion/.

The CASCB report 2022 is out. You find it here.

Today we presented our poster Neural Puppeteer at the 16th Asian Conference on Computer Vision (ACCV2022). The presentation was held virtually from Konstanz.

Code available for our unsupervised framework where we propagate poses and masks through a video: https://github.com/urs-waldmann/improving-unsupervised-label-propagation.

Check out our introduction video to Neural Puppeteer, that we will present at the 16th Asian Conference on Computer Vision (ACCV2022) in Macau SAR, China on December 7, 2022.