Michał Daniłowicz

2026

1. TRETS

   SIRA: scaled-integer range analysis for optimizing FPGA dataflow neural network accelerators

   Yaman Umuroglu, Christoph Berganski, Felix Jentzsch, Michal Danilowicz, Tomasz Kryjak, and 6 more authors

   ACM Transactions on Reconfigurable Technology and Systems, 2026

   Abstract Collaboration DOI arXiv Bib

   • Computer Engineering Group
     Paderborn University, Germany
   • AMD Research & Advanced Development
   • Computer Architecture Laboratory (CAL)
     Norwegian University of Science and Technology (NTNU), Norway

   All our collaborations

   While neural network quantization effectively reduces the cost of matrix multiplications, aggressive quantization can expose non-matrix-multiply operations as significant performance and resource bottlenecks on embedded systems. Addressing such bottlenecks requires a comprehensive approach to tailoring the precision across operations in the inference computation. To this end, we introduce scaled-integer range analysis ( SIRA ), a static analysis technique employing interval arithmetic to determine the range, scale, and bias for tensors in quantized neural networks. We show how this information can be exploited to reduce the resource footprint of FPGA dataflow neural network accelerators via tailored bitwidth adaptation for accumulators and downstream operations, aggregation of scales and biases, and conversion of consecutive elementwise operations to thresholding operations. We integrate SIRA -driven optimizations into the open source FINN framework, then evaluate their effectiveness across a range of quantized neural network workloads and compare implementation alternatives for non-matrix-multiply operations. We demonstrate an average reduction of 17% for LUTs, 66% for DSPs, and 22% for accumulator bitwidths with SIRA optimizations, providing detailed benchmark analysis and analytical models to guide the implementation style for non-matrix layers. Finally, we open source SIRA to facilitate community exploration of its benefits across various applications and hardware platforms.

   @article{badap-agh:168428,
     title = {SIRA: scaled-integer range analysis for optimizing FPGA dataflow neural network accelerators},
     author = {Umuroglu, Yaman and Berganski, Christoph and Jentzsch, Felix and Danilowicz, Michal and Kryjak, Tomasz and Bezaitis, Charalampos and Sj\"{a}lander, Magnus and Colbert, Ian and Preusser, Thomas and Petri-Koenig, Jakoba and Blott, Michaela},
     year = {2026},
     journal = {ACM Transactions on Reconfigurable Technology and Systems},
     series = {ACM Transactions on Reconfigurable Technology and Systems},
     volume = {19},
     number = {2},
     pages = {23},
     doi = {10.1145/3807510},
     issn = {1936-7406},
     collection = {ACM Transactions on Reconfigurable Technology and Systems},
   }

2025

1. ARC

   Real-time multi-object tracking using YOLOv8 and SORT on a SoC FPGA

   Michal Danilowicz and Tomasz Kryjak

   In Applied Reconfigurable Computing : architectures, tools, and applications : 21st international symposium, ARC 2025 : Seville, Spain, April 9–11, 2025 : proceedings, 2025

   Abstract DOI arXiv Bib

   Multi-object tracking (MOT) is one of the most important problems in computer vision and a key component of any vision-based perception system used in advanced autonomous mobile robotics. Therefore, its implementation on low-power and real-time embedded platforms is highly desirable. Modern MOT algorithms should be able to track objects of a given class (e.g. people or vehicles). In addition, the number of objects to be tracked is not known in advance, and they may appear and disappear at any time, as well as be obscured. For these reasons, the most popular and successful approaches have recently been based on the tracking paradigm. Therefore, the presence of a high quality object detector is essential, which in practice accounts for the vast majority of the computational and memory complexity of the whole MOT system. In this paper, we propose an FPGA (Field-Programmable Gate Array) implementation of an embedded MOT system based on a quantized YOLOv8 detector and the SORT (Simple Online Realtime Tracker) tracker. We use a modified version of the FINN framework to utilize external memory for model parameters and to support operations necessary required by YOLOv8. We discuss the evaluation of detection and tracking performance using the COCO and MOT15 datasets, where we achieve 0.21 mAP and 38.9 MOTA respectively. As the computational platform, we use an MPSoC system (Zynq UltraScale+ device from AMD/Xilinx) where the detector is deployed in reprogrammable logic and the tracking algorithm is implemented in the processor system.

   @inproceedings{badap-agh:159463,
     title = {Real-time multi-object tracking using YOLOv8 and SORT on a SoC FPGA},
     author = {Danilowicz, Michal and Kryjak, Tomasz},
     year = {2025},
     series = {Applied Reconfigurable Computing : architectures, tools, and applications : 21st international symposium, ARC~2025 : Seville, Spain, April 9–11,~2025 : proceedings},
     pages = {214–230},
     doi = {10.1007/978-3-031-87995-1_13},
     collection = {Applied Reconfigurable Computing : architectures, tools, and applications : 21st international symposium, ARC~2025 : Seville, Spain, April 9–11,~2025 : proceedings},
   }

2. DSD

   Hardware-aware feature extraction quantisation for real-time visual odometry on FPGA platforms

   Mateusz Wasala, Mateusz Smolarczyk, Michal Danilowicz, and Tomasz Kryjak

   In DSD 2025 : 28th euromicro conference on Digital System Design : 10–12 September 2025, Salerno, Italy, 2025

   Abstract DOI arXiv Bib

   Accurate position estimation is essential for modern navigation systems deployed in autonomous platforms, including ground vehicles, marine vessels, and aerial drones. In this context, Visual Simultaneous Localisation and Mapping (VSLAM) which includes Visual Odometry - relies heavily on the reliable extraction of salient feature points from the visual input data. In this work, we propose an embedded implementation of an unsupervised architecture capable of detecting and describing feature points. It is based on a quantised SuperPoint convolutional neural network. Our objective is to minimise the computational demands of the model while preserving high detection quality, thus facilitating efficient deployment on platforms with limited resources, such as mobile or embedded systems. We implemented the solution on an FPGA System-on-Chip (SoC) platform, specifically the AMD/Xilinx Zynq UltraScale+, where we evaluated the performance of Deep Learning Processing Units (DPUs) and we also used the Brevitas library and the FINN framework to perform model quantisation and hardware-aware optimisation. This allowed us to process 640 × 480 pixel images at up to 54 fps on an FPGA platform, outperforming state-of-the-art solutions in the field. We conducted experiments on the TUM dataset to demonstrate and discuss the impact of different quantisation techniques on the accuracy and performance of the model in a visual odometry task.

   @inproceedings{badap-agh:165139,
     title = {Hardware-aware feature extraction quantisation for real-time visual odometry on FPGA platforms},
     author = {Wasala, Mateusz and Smolarczyk, Mateusz and Danilowicz, Michal and Kryjak, Tomasz},
     year = {2025},
     series = {DSD~2025 [Dokument elektroniczny] : 28th euromicro conference on Digital System Design : 10–12 September~2025, Salerno, Italy},
     pages = {539–548},
     doi = {10.1109/DSD67783.2025.00080},
     collection = {DSD~2025 [Dokument elektroniczny] : 28th euromicro conference on Digital System Design : 10–12 September~2025, Salerno, Italy},
   }

2022

1. JSPS

   Automotive perception system evaluation with reference data from a UAV’s camera using ArUco markers and DCNN

   Krzysztof Blachut, Michal Danilowicz, Hubert Szolc, Mateusz Wasala, Tomasz Kryjak, and 1 more author

   Journal of Signal Processing Systems for Signal, Image, and Video Technology, 2022

   Abstract DOI Bib

   Testing and evaluation of an automotive perception system is a complicated task which requires special equipment and infrastructure. To compute key performance indicators and compare the results with real-world situation, some additional sensors and manual data labelling are often required. In this article, we propose a different approach, which is based on a UAV equipped with a 4K camera flying above a test track. Two computer vision methods are used to precisely determine the positions of the objects around the car – one based on ArUco markers and the other on a DCNN (we provide the algorithms used on GitHub). The detections are then correlated with the perception system readings. For the static and dynamic experiments, the differences between various systems are mostly below 0.5 m. The results of the experiments performed indicate that this approach could be an interesting alternative to existing evaluation solutions.

   @article{badap-agh:141200,
     title = {Automotive perception system evaluation with reference data from a UAV's camera using ArUco markers and DCNN},
     author = {Blachut, Krzysztof and Danilowicz, Michal and Szolc, Hubert and Wasala, Mateusz and Kryjak, Tomasz and Komorkiewicz, Mateusz},
     year = {2022},
     journal = {Journal of Signal Processing Systems for Signal, Image, and Video Technology},
     series = {Journal of Signal Processing Systems for Signal, Image, and Video Technology},
     volume = {94},
     number = {7},
     pages = {675},
     doi = {10.1007/s11265-021-01734-3},
     issn = {1939-8018},
     collection = {Journal of Signal Processing Systems for Signal, Image, and Video Technology},
   }

2. JLPEA

   Embedded object detection with custom LittleNet, FINN and vitis AI DCNN accelerators

   Michal Machura, Michal Danilowicz, and Tomasz Kryjak

   Journal of Low Power Electronics and Applications, 2022

   Abstract DOI Bib

   Object detection is an essential component of many systems used, for example, in advanced driver assistance systems (ADAS) or advanced video surveillance systems (AVSS). Currently, the highest detection accuracy is achieved by solutions using deep convolutional neural networks (DCNN). Unfortunately, these come at the cost of a high computational complexity; hence, the work on the widely understood acceleration of these algorithms is very important and timely. In this work, we compare three different DCNN hardware accelerator implementation methods: coarse-grained (a custom accelerator called LittleNet), fine-grained (FINN) and sequential (Vitis AI). We evaluate the approaches in terms of object detection accuracy, throughput and energy usage on the VOT and VTB datasets. We also present the limitations of each of the methods considered. We describe the whole process of DNNs implementation, including architecture design, training, quantisation and hardware implementation. We used two custom DNN architectures to obtain a higher accuracy, higher throughput and lower energy consumption. The first was implemented in SystemVerilog and the second with the FINN tool from AMD Xilinx. Next, both approaches were compared with the Vitis AI tool from AMD Xilinx. The final implementations were tested on the Avnet Ultra96-V2 development board with the Zynq UltraScale+ MPSoC ZCU3EG device. For two different DNNs architectures, we achieved a throughput of 196 fps for our custom accelerator and 111 fps for FINN. The same networks implemented with Vitis AI achieved 123.3 fps and 53.3 fps, respectively.

   @article{badap-agh:140225,
     title = {Embedded object detection with custom LittleNet, FINN and vitis AI DCNN accelerators},
     author = {Machura, Michal and Danilowicz, Michal and Kryjak, Tomasz},
     year = {2022},
     journal = {Journal of Low Power Electronics and Applications},
     series = {Journal of Low Power Electronics and Applications},
     volume = {12},
     number = {2},
     pages = {30},
     doi = {10.3390/jlpea12020030},
     issn = {2079-9268},
     collection = {Journal of Low Power Electronics and Applications},
   }

3. ARC

   Real-time embedded object tracking with discriminative correlation filters using convolutional features

   Michal Danilowicz and Tomasz Kryjak

   In Applied Reconfigurable Computing : architectures, tools, and applications : 18th international symposium, ARC 2022 : virtual event : September 19–20, 2022 : proceedings, 2022

   Abstract DOI Bib

   Object tracking is an essential element of visual perception systems. It is used in advanced video surveillance systems (AVSS), autonomous vehicles, robotics, and many more. For applications such as autonomous robots, the system must be implemented on some embedded platform with limited computing performance and power. Furthermore, sufficiently fast response is required from the tracking system in order to perform some real-time tasks. Discriminative Correlation Filter (DCF) based tracking algorithms are popular for such applications, as they offer state-of-the-art performance while not being too computationally complex. In this paper, an FPGA implementation of the DCF tracking algorithm using convolutional features is presented. The ZCU104 board is used as a platform, and the performance is evaluated on the VOT2015 dataset. In contrast to other implementations that use HOG (Histogram of Oriented Gradients) features, this implementation achieves better results for 64×64 filter size while being able to potentially operate at higher speeds (over 467 fps per scale).

   @inproceedings{badap-agh:143788,
     title = {Real-time embedded object tracking with discriminative correlation filters using convolutional features},
     author = {Danilowicz, Michal and Kryjak, Tomasz},
     year = {2022},
     series = {Applied Reconfigurable Computing : architectures, tools, and applications : 18th~international symposium, ARC~2022 : virtual event : September 19–20, 2022 : proceedings},
     pages = {166–180},
     doi = {10.1007/978-3-031-19983-7_12},
     collection = {Applied Reconfigurable Computing : architectures, tools, and applications : 18th~international symposium, ARC~2022 : virtual event : September 19–20, 2022 : proceedings},
   }

4. Book chapter Preludium
   

   Systemy wizyjne w zastosowaniach przemyslowych

   Marek Gorgon, Maciej Aleksandrowicz, Krzysztof Blachut, Andrzej Brodzicki, Zbigniew Bublinski, and 20 more authors

   In Nauka – technika – technologia : seria wydawnicza AGH, 2022

   Abstract DOI Bib

   Streszczenie: Systemy wizyjne to dynamicznie rozwijająca się dziedzina nauki i techniki.W niniejszym artykule przedstawiono przegląd zagadnień, nad którymi pracuje zespół Laboratorium Systemów Wizyjnych.Obejmuje on systemy wizyjne dla pojazdów autonomicznych, aplikacje analizy obrazów medycznych, systemy wspomagania szkolenia pracowników, analizę obrazów termowizyjnych, implementacje i akceleracje algorytmów wizyjnych w układach SoC FPGA i GPU, przetwarzanie strumienia wizyjnego w rozdzielczości 4K, a także analizę sygnałów z wykorzystaniem głębokich sieci neuronowych i współpracę przy eksperymencie ALICE w CERN-ie.

   @inbook{badap-agh:141409,
     title = {Systemy wizyjne w zastosowaniach przemyslowych},
     author = {Gorgon, Marek and Aleksandrowicz, Maciej and Blachut, Krzysztof and Brodzicki, Andrzej and Bublinski, Zbigniew and Cyba, Artur and Danilowicz, Michal and Glowacz, Adam and Jablonski, Miroslaw and Jaworek-Korjakowska, Joanna and Kostuch, Aleksander and Kowalczyk, Marcin and Kryjak, Tomasz and Kucharski, Dariusz and Lis, Konrad and Machura, Michal and Mikrut, Zbigniew and Pawlik, Piotr and Piekarski, Michal and Przewlocka-Rus, Dominika and Lis, Joanna and Szolc, Hubert and Wasala, Mateusz and Wojcicka, Anna Malgorzata and Wzorek, Piotr},
     year = {2022},
     projects = {Preludium-Kowalczyk},
     series = {Nauka – technika – technologia : seria wydawnicza AGH},
     volume = {T.~5},
     pages = {17–31},
     doi = {10.7494/978-83-67427-00-5_1},
     collection = {Nauka – technika – technologia : seria wydawnicza AGH},
   }

2021

1. DASIP

   Automotive perception system evaluation with reference data obtained by a UAV

   Krzysztof Blachut, Michal Danilowicz, Hubert Szolc, Mateusz Wasala, Tomasz Kryjak, and 2 more authors

   In DASIP 2021 : workshop on Design and Architectures for Signal and Image Processing - 14th edition : January 18–20 2021, Budapest, Hungary : proceedings, 2021

   Abstract DOI Bib

   Testing and evaluation of an automotive perception system is a complicated task which requires special equipment and infrastructure. To compute key performance indicators and compare the results with the real-world situation, some additional sensors and manual data labelling are often required. In this article, we propose a different approach, which is based on a UAV equipped with a 4K camera flying above the test track. Thanks to the synchronisation of the sensors between the tested vehicle and the UAV, it is possible to precisely determine the positions of the objects around the car and correlate them with the perception system readings. The performed experiments indicate that this approach could be an interesting alternative to the existing evaluation solutions.

   @inproceedings{badap-agh:132409,
     title = {Automotive perception system evaluation with reference data obtained by a UAV},
     author = {Blachut, Krzysztof and Danilowicz, Michal and Szolc, Hubert and Wasala, Mateusz and Kryjak, Tomasz and Pankiewicz, Nikodem and Komorkiewicz, Mateusz},
     year = {2021},
     series = {DASIP~2021 [Dokument elektroniczny] : workshop on Design and Architectures for Signal and Image Processing - 14th~edition : January 18–20 2021, Budapest, Hungary : proceedings},
     pages = {10–18},
     doi = {10.1145/3441110.3441151},
     collection = {DASIP~2021 [Dokument elektroniczny] : workshop on Design and Architectures for Signal and Image Processing - 14th~edition : January 18–20 2021, Budapest, Hungary : proceedings},
   }