Muhammet Balcilar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Muhammet Balcilar has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Mehmet Fatih Amasyali
Yildiz Technical University
4 shared publications - Erkan Uslu
Yildiz Technical University
3 shared publications Sirma Yavuz
Bilgisayar Mühendisliği Bölümü, Yıldız Teknik Üniversitesi, İstanbul, Türkiye
2 shared publications- A. Coskun Sonmez
Istanbul Technical University
1 shared publication Bedir Yilmaz
Bilgisayar Mühendisliği Bölümü, Iğdır Üniversitesi, Türkiye
1 shared publication- Bharath Bhushan Damodaran
UMR 6074 Université de Bretagne Sud IRISA CNRS France
1 shared publication
Latest Publications
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 34 (10577149) , 306-319
Exploiting Latent Properties to Optimize Neural Codecs
End-to-end image and video codecs are becoming increasingly competitive, compared to traditional compression techniques that have been developed through decades of manual engineering efforts. These trainable codecs have many advantages over traditional techniques, such as their straightforward adaptation to perceptual distortion metrics and high performance in specific fields thanks to their learning ability. However, current state-of-the-art neural codecs do not fully exploit the benefits of vector quantization and the existence of the entropy gradient in decoding devices. In this paper, we propose to leverage these two properties (vector quantization and entropy gradient) to improve the performance of off-the-shelf codecs. Firstly, we demonstrate that using non-uniform scalar quantization cannot improve performance over uniform quantization. We thus suggest using predefined optimal uniform vector quantization to improve performance. Secondly, we show that the entropy gradient, available at the decoder, is correlated with the reconstruction error gradient, which is not available at the decoder. We therefore use the former as a proxy to enhance compression performance. Our experimental results show that these approaches save between 1 to 3% of the rate for the same quality across various pre-trained methods. In addition, the entropy gradient based solution improves traditional codec performance significantly as well.
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 34 (10577149) , 306-319
Exploiting Latent Properties to Optimize Neural Codecs
End-to-end image and video codecs are becoming increasingly competitive, compared to traditional compression techniques that have been developed through decades of manual engineering efforts. These trainable codecs have many advantages over traditional techniques, such as their straightforward adaptation to perceptual distortion metrics and high performance in specific fields thanks to their learning ability. However, current state-of-the-art neural codecs do not fully exploit the benefits of vector quantization and the existence of the entropy gradient in decoding devices. In this paper, we propose to leverage these two properties (vector quantization and entropy gradient) to improve the performance of off-the-shelf codecs. Firstly, we demonstrate that using non-uniform scalar quantization cannot improve performance over uniform quantization. We thus suggest using predefined optimal uniform vector quantization to improve performance. Secondly, we show that the entropy gradient, available at the decoder, is correlated with the reconstruction error gradient, which is not available at the decoder. We therefore use the former as a proxy to enhance compression performance. Our experimental results show that these approaches save between 1 to 3% of the rate for the same quality across various pre-trained methods. In addition, the entropy gradient based solution improves traditional codec performance significantly as well.
- 1 January 2017
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 87, 66-80
R-SLAM: Resilient localization and mapping in challenging environments
- 1 January 2017
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 87, 66-80
R-SLAM: Resilient localization and mapping in challenging environments
- 1 May 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1513-1516
Deformable part model and deep learning comparison on victim detection
Object detection problem is a considerable research field that is being developed through continuous research. Simulated victims (dolls) detection performances of 2 different methods are given in the scope of this work. While deformable part model method is performing high accuracy and speed to detect object, with growing and remarkable popularity, deep learning method is noteworthy with higher performance results. This work mentions about the advantages and disadvantages of both methods and gives experimental results on a simulated environment.
- 1 May 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1513-1516
Deformable part model and deep learning comparison on victim detection
Object detection problem is a considerable research field that is being developed through continuous research. Simulated victims (dolls) detection performances of 2 different methods are given in the scope of this work. While deformable part model method is performing high accuracy and speed to detect object, with growing and remarkable popularity, deep learning method is noteworthy with higher performance results. This work mentions about the advantages and disadvantages of both methods and gives experimental results on a simulated environment.
- 1 May 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 193-196
Impact of artificial dataset enlargement on performance of deformable part models
There is a remarkable body of work for increasing the performance of Deformable Part Models (DPM), which is one of the most popular algorithms that are being used for object detection from digital images. The contribution that has been made to the object detection performance of the DPM algorithm via usage of larger datasets that has been created via production of artificial images from original images has been examined in this study. Various artificial dataset enlargement techniques that require no additional effort of labeling or data gathering have been compared on INRIA dataset and positive impact of artificial dataset enlargement has been observed. An increase on object detection performance has been noted both on the original INRIA dataset and its subsets.
- 1 May 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 193-196
Impact of artificial dataset enlargement on performance of deformable part models
There is a remarkable body of work for increasing the performance of Deformable Part Models (DPM), which is one of the most popular algorithms that are being used for object detection from digital images. The contribution that has been made to the object detection performance of the DPM algorithm via usage of larger datasets that has been created via production of artificial images from original images has been examined in this study. Various artificial dataset enlargement techniques that require no additional effort of labeling or data gathering have been compared on INRIA dataset and positive impact of artificial dataset enlargement has been observed. An increase on object detection performance has been noted both on the original INRIA dataset and its subsets.
- 15 March 2015
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 10 (1) , 85-92
Background estimation method with incremental iterative Re-weighted least squares
- 15 March 2015
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 10 (1) , 85-92
Background estimation method with incremental iterative Re-weighted least squares
- Open Access
- 1 January 2014
- Published by Springer Nature
- p. 316-325
Routing with Dijkstra in Mobile Ad-Hoc Networks
- 1 January 2014
- Published by Springer Nature
- p. 316-325
Open AccessRouting with Dijkstra in Mobile Ad-Hoc Networks