Zobrazeno 1 - 10
of 20
pro vyhledávání: '"Przewiȩźlikowski, Marcin"'
Autor:
Przewięźlikowski, Marcin, Balestriero, Randall, Jasiński, Wojciech, Śmieja, Marek, Zieliński, Bartosz
Masked Image Modeling (MIM) has emerged as a popular method for Self-Supervised Learning (SSL) of visual representations. However, for high-level perception tasks, MIM-pretrained models offer lower out-of-the-box representation quality than the Joint
Externí odkaz:
http://arxiv.org/abs/2412.03215
Collaborative self-supervised learning has recently become feasible in highly distributed environments by dividing the network layers between client devices and a central server. However, state-of-the-art methods, such as MocoSFL, are optimized for n
Externí odkaz:
http://arxiv.org/abs/2406.08267
Autor:
Batorski, Paweł, Malarz, Dawid, Przewięźlikowski, Marcin, Mazur, Marcin, Tadeja, Sławomir, Spurek, Przemysław
Neural radiance fields (NeRFs) are a widely accepted standard for synthesizing new 3D object views from a small number of base images. However, NeRFs have limited generalization properties, which means that we need to use significant computational re
Externí odkaz:
http://arxiv.org/abs/2402.01524
Autor:
Szatkowski, Filip, Pyla, Mateusz, Przewięźlikowski, Marcin, Cygert, Sebastian, Twardowski, Bartłomiej, Trzciński, Tomasz
In this work, we investigate exemplar-free class incremental learning (CIL) with knowledge distillation (KD) as a regularization strategy, aiming to prevent forgetting. KD-based methods are successfully used in CIL, but they often struggle to regular
Externí odkaz:
http://arxiv.org/abs/2308.09544
Autor:
Przewięźlikowski, Marcin, Pyla, Mateusz, Zieliński, Bartosz, Twardowski, Bartłomiej, Tabor, Jacek, Śmieja, Marek
Publikováno v:
Knowledge-Based Systems, 2024, 112572, ISSN 0950-7051
Self-supervised learning (SSL) is a powerful technique for learning from unlabeled data. By learning to remain invariant to applied data augmentations, methods such as SimCLR and MoCo can reach quality on par with supervised approaches. However, this
Externí odkaz:
http://arxiv.org/abs/2306.06082
Autor:
Borycki, Piotr, Kubacki, Piotr, Przewięźlikowski, Marcin, Kuśmierczyk, Tomasz, Tabor, Jacek, Spurek, Przemysław
The main goal of Few-Shot learning algorithms is to enable learning from small amounts of data. One of the most popular and elegant Few-Shot learning approaches is Model-Agnostic Meta-Learning (MAML). The main idea behind this method is to learn the
Externí odkaz:
http://arxiv.org/abs/2210.02796
Autor:
Przewięźlikowski, Marcin, Pyla, Mateusz, Zieliński, Bartosz, Twardowski, Bartłomiej, Tabor, Jacek, Śmieja, Marek
Publikováno v:
In Knowledge-Based Systems 3 December 2024 305
Autor:
Sendera, Marcin, Przewięźlikowski, Marcin, Karanowski, Konrad, Zięba, Maciej, Tabor, Jacek, Spurek, Przemysław
Few-shot models aim at making predictions using a minimal number of labeled examples from a given task. The main challenge in this area is the one-shot setting where only one element represents each class. We propose HyperShot - the fusion of kernels
Externí odkaz:
http://arxiv.org/abs/2203.11378
Autor:
Przewięźlikowski, Marcin, Przybysz, Przemysław, Tabor, Jacek, Zięba, Maciej, Spurek, Przemysław
Publikováno v:
In Neurocomputing 14 September 2024 598
Processing of missing data by modern neural networks, such as CNNs, remains a fundamental, yet unsolved challenge, which naturally arises in many practical applications, like image inpainting or autonomous vehicles and robots. While imputation-based
Externí odkaz:
http://arxiv.org/abs/2110.14010