Adjusting Pretrained Backbones for Performativity

1 code implementation • 6 Oct 2024 • Berker Demirel, Lingjing Kong, Kun Zhang, Theofanis Karaletsos, Celestine Mendler-Dünner, Francesco Locatello

With the widespread deployment of deep learning models, they influence their environment in various ways.

Benchmarking Deep Learning

Look Around and Find Out: OOD Detection with Relative Angles

1 code implementation • 6 Oct 2024 • Berker Demirel, Marco Fumero, Francesco Locatello

In this work, we propose a novel angle-based metric for OOD detection that is computed relative to the in-distribution structure.

ADRMX: Additive Disentanglement of Domain Features with Remix Loss

1 code implementation • 12 Aug 2023 • Berker Demirel, Erchan Aptoula, Huseyin Ozkan

To this end, most of existing studies focus on extracting domain invariant features across the available source domains in order to mitigate the effects of inter-domain distributional changes.

Data Augmentation Disentanglement +1

DECOMPL: Decompositional Learning with Attention Pooling for Group Activity Recognition from a Single Volleyball Image

1 code implementation • 11 Mar 2023 • Berker Demirel, Huseyin Ozkan

To that end, we propose a novel GAR technique for volleyball videos, DECOMPL, which consists of two complementary branches.

Group Activity Recognition Optical Flow Estimation