Keuper Labs

Research in #ComputerVision , #MachineLearning, #HPC and #Geophysics

[Keuper Labs @KeuperLabs]

We are now on #Bluesky : @keuper-labs.bsky.social

[Paul Gavrikov @PaulGavrikov]

Today at #NeurIPS2024 Interpretable AI Workshop! Poster sessions are at 10am and 4pm. Big thanks to @sukrutrao for bringing our poster and @RutaBinky for presenting!!

[Janis Keuper 🇮🇱 @JanisKeuper]

Congratulations to Mario Fernandez for his successful defense in Paris at @ENS_ULM
- "Towards deep learning for seismic processing". Joint supervision with Matthias Delescluse (ENS) @FraunhoferITWM
@KeuperLabs

[Janis Keuper 🇮🇱 @JanisKeuper]

New IJCV paper out: Reliable Evaluation of Attribution Maps in CNNs: A Perturbation-Based
Approach. Joint work with @FraunhoferITWM and @KeuperLabs https://t.co/RtyhwEB9I0

[Janis Keuper 🇮🇱 @JanisKeuper]

Congratulations to @PaulaHarder8 for her successful PhD defense. Joint work at @rptu_kl_ld @FraunhoferITWM and @KeuperLabs on „Physics-Constrained Deep Learning
for Accelerating Climate Modeling“

[Paul Gavrikov @PaulGavrikov]

"How Do Training Methods Influence the Utilization of Vision Models?" accepted at #NeurIPS2024 IAI.

Not all layers in a DNN contribute equally to the final prediction—some don’t contribute at all. In our ongoing study, we explore this in ImageNet classification by fixing training data and architecture and reveal that different training methods (like SSL and AT) can significantly impact which layers become “critical” to the prediction.

https://t.co/yQMZQH9xI2
w/ @shashankska @margret_keuper @JanisKeuper

[Yumeng Li @YumengLi_007]

🤗Code is out: https://t.co/8dJuNngN6n
🚀TL;DR: VSTAR generates longer videos with dynamic visual evolution in a single pass. No fine-tuning is needed!
🙌Check out our project page: https://t.co/5gcFe0uat7

Bill Beluch @margret_keuper @isDanZhang @anna_khoreva @Bosch_AI ❤️

[Julia Grabinski @JuliaGrabinski]

Honored to be recognized as an Outstanding Reviewer for @eccvconf. 😊 Happy to contribute and work in this community!

[Shashank Agnihotri @shashankska]

Improving Feature Stability during Upsampling -- Spectral Artifacts and the Importance of Spatial Context, Accepted at @eccvconf 2024, a joint work with @JuliaGrabinski and @margret_keuper. Affiliations: @dwsunima @FraunhoferITWM @cvml_mpiinf @KeuperLabs

[Or Litany @orlitany]

Last week, we had the pleasure of virtually hosting @JuliaGrabinski in my group at the Technion. She shared her clever and creative use of neural fields for representing arbitrarily large filters in CNNs. Thanks for the insightful talk!

[Computer Vision and Machine Learning - MPI-INF @cvml_mpiinf]

"Implicit Representations for Constrained Image Segmentation" by Jan Philipp Schneider, @_Mishal_Fatima, @jovita_lukasik, Andreas Kolb, @margret_keuper, and Michael Moeller, in collaboration with @UniSiegen and @dwsunima.

[Computer Vision and Machine Learning - MPI-INF @cvml_mpiinf]

Paper: https://t.co/hRPdbID0M2
Project Page: https://t.co/t964mTfklJ
Poster Session: Thu 25 Jul 4:30 a.m. PDT — 6 a.m. PDT in Hall C 4-9 #809.

[Computer Vision and Machine Learning - MPI-INF @cvml_mpiinf]

"CosPGD: An efficient white-box adversarial attack for pixel-wise prediction tasks" by @shashankska, @jung_vision, @margret_keuper, in collaboration with @dwsunima.

https://t.co/TWGEqA3Bn3

[Shashank Agnihotri @shashankska]

Interested in a quick summary of our @icmlconf 2024 paper, CosPGD: An efficient white-box adversarial attack for pixel-wise prediction tasks?

Please watch this short video: https://t.co/eRlkCDUIk5

For more details, please checkout our project page: https://t.co/CbJdAQTRDh

[Paul Gavrikov @PaulGavrikov]

Attending #CVPR2024 in #Seattle was an incredible experience! 🌟It was wonderful to reconnect with familiar faces and friends, meet new people (including @drfeifei ), present my works, and immerse myself in the latest trends in computer vision.

[Paul Gavrikov @PaulGavrikov]

Presenting today at #CVPR2024!

Main Poster Session 5
Friday, June 21, Poster 253, Arch 4A-E, 10:30AM

[Margret Keuper @margret_keuper]

Happy to announce that our paper on increasing model robustness with only a simple input transformation is accepted a TMLR. Joint work @dwsunima, @cvml_mpiinf and @UniSiegen! https://t.co/nmz17s9mFD

[Paul Gavrikov @PaulGavrikov]

Are you in Seattle for #CVPR2024? Come and visit our poster today at 9:40 AM at the @MMFMWorkshop. We are looking forward to seeing you!

[Paul Gavrikov @PaulGavrikov]

On my way to #CVPR24 in Seattle to present two papers:

👉 Can Biases in ImageNet Models Explain Generalization?
👉 Are Vision Language Models Texture or Shape Biased and Can We Steer Them?

Please stop by and say hi if you are in Seattle!

[Katharina Prasse @KatharinaP19256]

On my way to #AISCICOMM24 in Zurich to present our joint research on manipulation in climate change images.
@IsaacBr45419303 @dwsunima @margret_keuper @SteWalter @TU_Muenchen @KeuperLabs