Attila Lengyel

I’m interested in both fundamental deep learning research as well as developing computer vision technology to help tackle today’s global challenges. My main research interests include representation learning, equivariant CNNs and efficient color encoding in neural networks.

Besides that, I also enjoy (kite)surfing, hiking and playing the guitar.

If you’re interested to collaborate, feel free to drop me an email.

news

May 9, 2022	I’ll be joining Google Zürich as a Research Intern, working on robust learned descriptors.
Mar 26, 2022	Our 3rd workshop on Visual Inductive Priors has been accepted to ECCV 2022.
Dec 16, 2021	I’m reviewing for CVPR 2022.
Jul 23, 2021	Our paper “Zero-Shot Day-Night Domain Adaptation with a Physics Prior” got accepted to ICCV 2021 as an oral presentation.
May 20, 2021	Our paper on symmetries in group equivariant convolutions got accepted to ICIP 2021.

selected publications

arXiv

VIPriors 2: Visual Inductive Priors for Data-Efficient Deep Learning Challenges

Attila Lengyel, Robert-Jan Bruintjes, Marcos Baptista-Rios, Osman Semih Kayhan, Davide Zambrano, Nergis Tomen, and Jan Gemert

2022

Abs HTML PDF

The second edition of the "VIPriors: Visual Inductive Priors for Data-Efficient Deep Learning" challenges featured five data-impaired challenges, where models are trained from scratch on a reduced number of training samples for various key computer vision tasks. To encourage new and creative ideas on incorporating relevant inductive biases to improve the data efficiency of deep learning models, we prohibited the use of pre-trained checkpoints and other transfer learning techniques. The provided baselines are outperformed by a large margin in all five challenges, mainly thanks to extensive data augmentation policies, model ensembling, and data efficient network architectures.
ICCV 2021

Zero-Shot Day-Night Domain Adaptation With a Physics Prior

Attila Lengyel, Sourav Garg, Michael Milford, and Jan Gemert

2021

Abs HTML PDF

We explore the zero-shot setting for day-night domain adaptation. The traditional domain adaptation setting is to train on one domain and adapt to the target domain by exploiting unlabeled data samples from the test set. As gathering relevant test data is expensive and sometimes even impossible, we remove any reliance on test data imagery and instead exploit a visual inductive prior derived from physics-based reflection models for domain adaptation. We cast a number of color invariant edge detectors as trainable layers in a convolutional neural network and evaluate their robustness to illumination changes. We show that the color invariant layer reduces the day-night distribution shift in feature map activations throughout the network. We demonstrate improved performance for zero-shot day to night domain adaptation on both synthetic as well as natural datasets in various tasks, including classification, segmentation and place recognition.
ICIP 2021

Exploiting Learned Symmetries in Group Equivariant Convolutions

Attila Lengyel, and Jan Gemert

2021

Abs HTML PDF

Group Equivariant Convolutions (GConvs) enable convolutional neural networks to be equivariant to various transformation groups, but at an additional parameter and compute cost. We investigate the filter parameters learned by GConvs and find certain conditions under which they become highly redundant. We show that GConvs can be efficiently decomposed into depthwise separable convolutions while preserving equivariance properties and demonstrate improved performance and data efficiency on two datasets. All code is publicly available at github.com/Attila94/SepGroupPy.