1 code implementation • ICML Workshop LifelongML 2020 • Kishan Parshotam, Mert Kilickaya
Our extensive experiments on three large-scale datasets, using two different architectures for five different continual learning methods, reveal that Normalised cross-entropy and synthetic transfer leads to less forgetting in existing techniques.
1 code implementation • 4 Sep 2023 • Fangqin Zhou, Mert Kilickaya, Joaquin Vanschoren
Hyperspectral image classification is gaining popularity for high-precision vision tasks in remote sensing, thanks to their ability to capture visual information available in a wide continuum of spectra.
Ranked #1 on Hyperspectral Image Classification on Houston (OA@15perclass metric)
1 code implementation • 1 Feb 2023 • Mert Kilickaya, Joost Van de Weijer, Yuki M. Asano
The current dominant paradigm when building a machine learning model is to iterate over a dataset over and over until convergence.
no code implementations • EACL 2017 • Mert Kilickaya, Aykut Erdem, Nazli Ikizler-Cinbis, Erkut Erdem
The task of generating natural language descriptions from images has received a lot of attention in recent years.
no code implementations • 10 Jun 2020 • Mert Kilickaya, Arnold Smeulders
To that end, in this paper, we propose to diagnose rarity in HOI detection.
no code implementations • 17 Oct 2020 • Mert Kilickaya, Noureldien Hussein, Efstratios Gavves, Arnold Smeulders
Our experiments show that SSC leads to an important increase in interaction recognition performance, while using much fewer parameters.
no code implementations • 27 Oct 2020 • Mert Kilickaya, Arnold W. M. Smeulders
The structure is in the form of a 2D composition that encodes the position and the category of the objects.
no code implementations • 1 Dec 2021 • Mert Kilickaya, Arnold Smeulders
iii) We evaluate Align-Former on HICO-DET [5] and V-COCO [13], and show that Align-Former outperforms existing image-level supervised HO-I detectors by a large margin (4. 71% mAP improvement from 16. 14% to 20. 85% on HICO-DET [5]).
no code implementations • 26 Jan 2023 • Mert Kilickaya, Joaquin Vanschoren
We propose VINIL, a self-incremental learner that can learn object instances sequentially, ii.
no code implementations • 23 Mar 2023 • Elif Ceren Gok Yildirim, Murat Onur Yildirim, Mert Kilickaya, Joaquin Vanschoren
Class-Incremental Learning updates a deep classifier with new categories while maintaining the previously observed class accuracy.
no code implementations • 20 Nov 2023 • Mert Kilickaya, Joaquin Vanschoren
This position paper outlines the potential of AutoML for incremental (continual) learning to encourage more research in this direction.