Tomasz Kryjak

Konrad Lis, Tomasz Kryjak

3D object detection from LiDAR sensor data is an important topic in the context of autonomous cars and drones. In this paper, we present the results of experiments on the impact of backbone selection of a deep convolutional neural network on detection accuracy and computation speed. We chose the PointPillars network, which is characterised by a simple architecture, high speed, and modularity that allows for easy expansion. During the experiments, we paid particular attention to the change in detection efficiency (measured by the mAP metric) and the total number of multiply-addition operations needed to process one point cloud. We tested 10 different convolutional neural network architectures that are widely used in image-based detection problems. For a backbone like MobilenetV1, we obtained an almost 4x speedup at the cost of a 1.13% decrease in mAP. On the other hand, for CSPDarknet we got an acceleration of more than 1.5x at an increase in mAP of 0.33%. We have thus demonstrated that it is possible to significantly speed up a 3D object detector in LiDAR point clouds with a small decrease in detection efficiency. This result can be used when PointPillars or similar algorithms are implemented in embedded systems, including SoC FPGAs. The code is available at https://github.com/vision-agh/pointpillars\_backbone.

Maciej Baczmanski, Robert Synoczek, Mateusz Wasala et al.

Object detection and segmentation are two core modules of an autonomous vehicle perception system. They should have high efficiency and low latency while reducing computational complexity. Currently, the most commonly used algorithms are based on deep neural networks, which guarantee high efficiency but require high-performance computing platforms. In the case of autonomous vehicles, i.e. cars, but also drones, it is necessary to use embedded platforms with limited computing power, which makes it difficult to meet the requirements described above. A reduction in the complexity of the network can be achieved by using an appropriate: architecture, representation (reduced numerical precision, quantisation, pruning), and computing platform. In this paper, we focus on the first factor - the use of so-called detection-segmentation networks as a component of a perception system. We considered the task of segmenting the drivable area and road markings in combination with the detection of selected objects (pedestrians, traffic lights, and obstacles). We compared the performance of three different architectures described in the literature: MultiTask V3, HybridNets, and YOLOP. We conducted the experiments on a custom dataset consisting of approximately 500 images of the drivable area and lane markings, and 250 images of detected objects. Of the three methods analysed, MultiTask V3 proved to be the best, achieving 99% mAP_50 for detection, 97% MIoU for drivable area segmentation, and 91% MIoU for lane segmentation, as well as 124 fps on the RTX 3060 graphics card. This architecture is a good solution for embedded perception systems for autonomous vehicles. The code is available at: https://github.com/vision-agh/MMAR_2023.

Konrad Lis, Tomasz Kryjak

Object detection in 3D is a crucial aspect in the context of autonomous vehicles and drones. However, prototyping detection algorithms is time-consuming and costly in terms of energy and environmental impact. To address these challenges, one can check the effectiveness of different models by training on a subset of the original training set. In this paper, we present a comparison of three algorithms for selecting such a subset - random sampling, random per class sampling, and our proposed MONSPeC (Maximum Object Number Sampling per Class). We provide empirical evidence for the superior effectiveness of random per class sampling and MONSPeC over basic random sampling. By replacing random sampling with one of the more efficient algorithms, the results obtained on the subset are more likely to transfer to the results on the entire dataset. The code is available at: https://github.com/vision-agh/monspec.

Kamil Jeziorek, Piotr Wzorek, Krzysztof Blachut et al.

As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a growing need for hardware-aware neural architectures that enable efficient, low-latency, and energy-conscious local processing. We present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on a SoC FPGA and evaluated on two open-source datasets. For classification task, our baseline floating-point model achieves 92.7% accuracy on SHD dataset - only 2.4% below the state of the art - while requiring over 10x and 67x fewer parameters. On SSC, our models achieve 66.9-71.0% accuracy. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by up to 19.3% while reducing resource usage and latency. For SSC, we report the first hardware-accelerated evaluation. We further demonstrate the first end-to-end FPGA implementation of event-audio keyword spotting, combining graph convolutional layers with recurrent sequence modelling. The system achieves up to 95% word-end detection accuracy, with only 10.53 microsecond latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS.

Kamil Jeziorek, Andrea Pinna, Tomasz Kryjak

Recent advances in event camera research emphasize processing data in its original sparse form, which allows the use of its unique features such as high temporal resolution, high dynamic range, low latency, and resistance to image blur. One promising approach for analyzing event data is through graph convolutional networks (GCNs). However, current research in this domain primarily focuses on optimizing computational costs, neglecting the associated memory costs. In this paper, we consider both factors together in order to achieve satisfying results and relatively low model complexity. For this purpose, we performed a comparative analysis of different graph convolution operations, considering factors such as execution time, the number of trainable model parameters, data format requirements, and training outcomes. Our results show a 450-fold reduction in the number of parameters for the feature extraction module and a 4.5-fold reduction in the size of the data representation while maintaining a classification accuracy of 52.3%, which is 6.3% higher compared to the operation used in state-of-the-art approaches. To further evaluate performance, we implemented the object detection architecture and evaluated its performance on the N-Caltech101 dataset. The results showed an accuracy of 53.7 % mAP@0.5 and reached an execution rate of 82 graphs per second.

Tomasz Kryjak

In recent years there has been a growing interest in event cameras, i.e. vision sensors that record changes in illumination independently for each pixel. This type of operation ensures that acquisition is possible in very adverse lighting conditions, both in low light and high dynamic range, and reduces average power consumption. In addition, the independent operation of each pixel results in low latency, which is desirable for robotic solutions. Nowadays, Field Programmable Gate Arrays (FPGAs), along with general-purpose processors (GPPs/CPUs) and programmable graphics processing units (GPUs), are popular architectures for implementing and accelerating computing tasks. In particular, their usefulness in the embedded vision domain has been repeatedly demonstrated over the past 30 years, where they have enabled fast data processing (even in real-time) and energy efficiency. Hence, the combination of event cameras and reconfigurable devices seems to be a good solution, especially in the context of energy-efficient real-time embedded systems. This paper gives an overview of the most important works, where FPGAs have been used in different contexts to process event data. It covers applications in the following areas: filtering, stereovision, optical flow, acceleration of AI-based algorithms (including spiking neural networks) for object classification, detection and tracking, and applications in robotics and inspection systems. Current trends and challenges for such systems are also discussed.

Sylwia Kuros, Tomasz Kryjak

Quantum Neural Networks (QNNs) are an emerging technology that can be used in many applications including computer vision. In this paper, we presented a traffic sign classification system implemented using a hybrid quantum-classical convolutional neural network. Experiments on the German Traffic Sign Recognition Benchmark dataset indicate that currently QNN do not outperform classical DCNN (Deep Convolutuional Neural Networks), yet still provide an accuracy of over 90% and are a definitely promising solution for advanced computer vision.

Mateusz Wasala, Tomasz Kryjak

Object detection is an essential component of many vision systems. For example, pedestrian detection is used in advanced driver assistance systems (ADAS) and advanced video surveillance systems (AVSS). Currently, most detectors use deep convolutional neural networks (e.g., the YOLO -- You Only Look Once -- family), which, however, due to their high computational complexity, are not able to process a very high-resolution video stream in real-time, especially within a limited energy budget. In this paper we present a hardware implementation of the well-known pedestrian detector with HOG (Histogram of Oriented Gradients) feature extraction and SVM (Support Vector Machine) classification. Our system running on AMD Xilinx Zynq UltraScale+ MPSoC (Multiprocessor System on Chip) device allows real-time processing of 4K resolution (UHD -- Ultra High Definition, 3840 x 2160 pixels) video for 60 frames per second. The system is capable of detecting a pedestrian in a single scale. The results obtained confirm the high suitability of reprogrammable devices in the real-time implementation of embedded vision systems.

Wiktor Matykiewicz, Piotr Wzorek, Kamil Jeziorek et al.

With the rapid growth of mobile robotics and embedded intelligence, there is an increasing demand for efficient on-device data processing on edge platforms. A promising research direction is the use of neuromorphic sensors inspired by human sensory systems, which generate sparse, event-based data encoding changes in the environment. In this work, we present the first end-to-end FPGA implementation of a keyword spotting system that integrates a Neuromorphic Auditory Sensor (NAS) and a graph neural network (GNN) on a single FPGA device, enabling real-time processing of raw audio data. The proposed architecture eliminates conventional signal preprocessing and operates directly on event-based audio streams. Leveraging a compute-near-memory network architecture, the system achieves efficient inference with low latency and low power consumption. Experimental results demonstrate an accuracy of 87.43% after quantization on the Google Speech Commands v2 dataset processed through the neuromorphic sensor, with end-to-end latency below 35 us and average power consumption of 1.12 W. The processed datasets, software models, and hardware modules are available at https://github.com/vision-agh/NAS-GNN-KWS.

Dominika Przewlocka-Rus, Tomasz Kryjak, Marek Gorgon

The performance of object detection systems in automotive solutions must be as high as possible, with minimal response time and, due to the often battery-powered operation, low energy consumption. When designing such solutions, we therefore face challenges typical for embedded vision systems: the problem of fitting algorithms of high memory and computational complexity into small low-power devices. In this paper we propose PowerYOLO - a mixed precision solution, which targets three essential elements of such application. First, we propose a system based on a Dynamic Vision Sensor (DVS), a novel sensor, that offers low power requirements and operates well in conditions with variable illumination. It is these features that may make event cameras a preferential choice over frame cameras in some applications. Second, to ensure high accuracy and low memory and computational complexity, we propose to use 4-bit width Powers-of-Two (PoT) quantisation for convolution weights of the YOLO detector, with all other parameters quantised linearly. Finally, we embrace from PoT scheme and replace multiplication with bit-shifting to increase the efficiency of hardware acceleration of such solution, with a special convolution-batch normalisation fusion scheme. The use of specific sensor with PoT quantisation and special batch normalisation fusion leads to a unique system with almost 8x reduction in memory complexity and vast computational simplifications, with relation to a standard approach. This efficient system achieves high accuracy of mAP 0.301 on the GEN1 DVS dataset, marking the new state-of-the-art for such compressed model.

Piotr Wzorek, Tomasz Kryjak

In recent years, event cameras (DVS - Dynamic Vision Sensors) have been used in vision systems as an alternative or supplement to traditional cameras. They are characterised by high dynamic range, high temporal resolution, low latency, and reliable performance in limited lighting conditions -- parameters that are particularly important in the context of advanced driver assistance systems (ADAS) and self-driving cars. In this work, we test whether these rather novel sensors can be applied to the popular task of traffic sign detection. To this end, we analyse different representations of the event data: event frame, event frequency, and the exponentially decaying time surface, and apply video frame reconstruction using a deep neural network called FireNet. We use the deep convolutional neural network YOLOv4 as a detector. For particular representations, we obtain a detection accuracy in the range of 86.9-88.9% mAP@0.5. The use of a fusion of the considered representations allows us to obtain a detector with higher accuracy of 89.9% mAP@0.5. In comparison, the detector for the frames reconstructed with FireNet is characterised by an accuracy of 72.67% mAP@0.5. The results obtained illustrate the potential of event cameras in automotive applications, either as standalone sensors or in close cooperation with typical frame-based cameras.

Marcin Kowalczyk, Tomasz Kryjak

Neuromorphic vision is a rapidly growing field with numerous applications in the perception systems of autonomous vehicles. Unfortunately, due to the sensors working principle, there is a significant amount of noise in the event stream. In this paper we present a novel algorithm based on an IIR filter matrix for filtering this type of noise and a hardware architecture that allows its acceleration using an SoC FPGA. Our method has a very good filtering efficiency for uncorrelated noise - over 99% of noisy events are removed. It has been tested for several event data sets with added random noise. We designed the hardware architecture in such a way as to reduce the utilisation of the FPGA's internal BRAM resources. This enabled a very low latency and a throughput of up to 385.8 MEPS million events per second.The proposed hardware architecture was verified in simulation and in hardware on the Xilinx Zynq Ultrascale+ MPSoC chip on the Mercury+ XU9 module with the Mercury+ ST1 base board.

Dominika Przewlocka-Rus, Tomasz Kryjak

Deep neural networks virtually dominate the domain of most modern vision systems, providing high performance at a cost of increased computational complexity.Since for those systems it is often required to operate both in real-time and with minimal energy consumption (e.g., for wearable devices or autonomous vehicles, edge Internet of Things (IoT), sensor networks), various network optimisation techniques are used, e.g., quantisation, pruning, or dedicated lightweight architectures. Due to the logarithmic distribution of weights in neural network layers, a method providing high performance with significant reduction in computational precision (for 4-bit weights and less) is the Power-of-Two (PoT) quantisation (and therefore also with a logarithmic distribution). This method introduces additional possibilities of replacing the typical for neural networks Multiply and ACcumulate (MAC -- performing, e.g., convolution operations) units, with more energy-efficient Bitshift and ACcumulate (BAC). In this paper, we show that a hardware neural network accelerator with PoT weights implemented on the Zynq UltraScale + MPSoC ZCU104 SoC FPGA can be at least $1.4x$ more energy efficient than the uniform quantisation version. To further reduce the actual power requirement by omitting part of the computation for zero weights, we also propose a new pruning method adapted to logarithmic quantisation.

Kamil Bialik, Marcin Kowalczyk, Krzysztof Blachut et al.

This paper proposes the use of an event camera as a component of a vision system that enables counting of fast-moving objects - in this case, falling corn grains. These type of cameras transmit information about the change in brightness of individual pixels and are characterised by low latency, no motion blur, correct operation in different lighting conditions, as well as very low power consumption. The proposed counting algorithm processes events in real time. The operation of the solution was demonstrated on a stand consisting of a chute with a vibrating feeder, which allowed the number of grains falling to be adjusted. The objective of the control system with a PID controller was to maintain a constant average number of falling objects. The proposed solution was subjected to a series of tests to determine the correctness of the developed method operation. On their basis, the validity of using an event camera to count small, fast-moving objects and the associated wide range of potential industrial applications can be confirmed.

Krzysztof Blachut, Tomasz Kryjak

In this paper we have addressed the implementation of the accumulation and projection of high-resolution event data stream (HD -1280 x 720 pixels) onto the image plane in FPGA devices. The results confirm the feasibility of this approach, but there are a number of challenges, limitations and trade-offs to be considered. The required hardware resources of selected data representations, such as binary frame, event frame, exponentially decaying time surface and event frequency, were compared with those available on several popular platforms from AMD Xilinx. The resulting event frames can be used for typical vision algorithms, such as object classification and detection, using both classical and deep neural network methods.

Maciej Baczmanski, Mateusz Wasala, Tomasz Kryjak

Perception and control systems for autonomous vehicles are an active area of scientific and industrial research. These solutions should be characterised by high efficiency in recognising obstacles and other environmental elements in different road conditions, real-time capability, and energy efficiency. Achieving such functionality requires an appropriate algorithm and a suitable computing platform. In this paper, we have used the MultiTaskV3 detection-segmentation network as the basis for a perception system that can perform both functionalities within a single architecture. It was appropriately trained, quantised, and implemented on the AMD Xilinx Kria KV260 Vision AI embedded platform. By using this device, it was possible to parallelise and accelerate the computations. Furthermore, the whole system consumes relatively little power compared to a CPU-based implementation (an average of 5 watts, compared to the minimum of 55 watts for weaker CPUs, and the small size (119mm x 140mm x 36mm) of the platform allows it to be used in devices where the amount of space available is limited. It also achieves an accuracy higher than 97% of the mAP (mean average precision) for object detection and above 90% of the mIoU (mean intersection over union) for image segmentation. The article also details the design of the Mecanum wheel vehicle, which was used to test the proposed solution in a mock-up city.

Mariusz Grabowski, Tomasz Kryjak

In this paper, we propose a real-time FPGA implementation of the Semi-Global Matching (SGM) stereo vision algorithm. The designed module supports a 4K/Ultra HD (3840 x 2160 pixels @ 30 frames per second) video stream in a 4 pixel per clock (ppc) format and a 64-pixel disparity range. The baseline SGM implementation had to be modified to process pixels in the 4ppc format and meet the timing constrains, however, our version provides results comparable to the original design. The solution has been positively evaluated on the Xilinx VC707 development board with a Virtex-7 FPGA device.

Hubert Szolc, Tomasz Kryjak

This paper presents a system for hardware-in-the-loop (HiL) simulation of unmanned aerial vehicle (UAV) control algorithms implemented on a heterogeneous SoC FPGA computing platforms. The AirSim simulator running on a PC and an Arty Z7 development board with a Zynq SoC chip from AMD Xilinx were used. Communication was carried out via a serial USB link. An application for autonomous landing on a specially marked landing strip was selected as a case study. A landing site detection algorithm was implemented on the Zynq SoC platform. This allowed processing a 1280 x 720 @ 60 fps video stream in real time. Performed tests showed that the system works correctly and there are no delays that could negatively affect the stability of the control. The proposed concept is characterised by relative simplicity and low implementation cost. At the same time, it can be applied to test various types of high-level perception and control algorithms for UAV implemented on embedded platforms. We provide the code developed on GitHub, which includes both Python scripts running on the PC and C code running on Arty Z7.

Kamil Jeziorek, Tomasz Kryjak

This paper presents a method for detection and recognition of traffic signs based on information extracted from an event camera. The solution used a FireNet deep convolutional neural network to reconstruct events into greyscale frames. Two YOLOv4 network models were trained, one based on greyscale images and the other on colour images. The best result was achieved for the model trained on the basis of greyscale images, achieving an efficiency of 87.03%.

Dominika Przewlocka-Rus, Tomasz Kryjak

Siamese trackers have been among the state-of-the-art solutions in each Visual Object Tracking (VOT) challenge over the past few years. However, with great accuracy comes great computational complexity: to achieve real-time processing, these trackers have to be massively parallelised and are usually run on high-end GPUs. Easy to implement, this approach is energy consuming, and thus cannot be used in many low-power applications. To overcome this, one can use energy-efficient embedded devices, such as heterogeneous platforms joining the ARM processor system with programmable logic (FPGA). In this work, we propose a hardware-software implementation of the well-known fully connected Siamese tracker (SiamFC). We have developed a quantised Siamese network for the FINN accelerator, using algorithm-accelerator co-design, and performed design space exploration to achieve the best efficiency-to-energy ratio (determined by FPS and used resources). For our network, running in the programmable logic part of the Zynq UltraScale+ MPSoC ZCU104, we achieved the processing of almost 50 frames-per-second with tracker accuracy on par with its floating point counterpart, as well as the original SiamFC network. The complete tracking system, implemented in ARM with the network accelerated on FPGA, achieves up to 17 fps. These results bring us towards bridging the gap between the highly accurate but energy-demanding algorithms and energy-efficient solutions ready to be used in low-power, edge systems.

Michal Filipkowski, Marcin Kowalczyk, Tomasz Kryjak

This paper presents a hardware architecture that implements the Contrast Maximization (CM) algorithm in Field-Programmable Gate Array (FPGA) resources for event-based vision systems. CM estimates motion parameters by maximizing the contrast of an Image of Warped Events (IWE) reconstructed from asynchronous event streams. Event-based vision sensors generate sparse data with high temporal resolution and low spatial redundancy, which makes them well suited for hardware processing. The deterministic, massively parallel structure of the FPGA is leveraged to design a deeply pipelined architecture capable of high-throughput, energy-efficient processing suitable for real-time embedded applications. This paper details the hardware modules responsible for event warping, contrast computation, and iterative optimization, discusses key implementation decisions, and presents the hardware-aware optimization method used in the design. Experimental results demonstrate a substantial speed and efficiency improvement over CPU- and GPU-based implementations, with motion parameter estimation executing over 200 times faster. To the best of our knowledge, this is the first hardware architecture enabling acceleration of CM algorithm computations. Its performance is evaluated in terms of processing speed, energy efficiency, and hardware resource utilization. The proposed design is validated using an event-based object tracking application. The results confirm that the architecture provides a solid foundation for real-time motion estimation in high-speed, low-power embedded systems.

Marcin Kowalczyk, Kamil Jeziorek, Tomasz Kryjak

Event-based sensors offer significant advantages over traditional frame-based cameras, especially in scenarios involving rapid motion or challenging lighting conditions. However, event data frequently suffers from considerable noise, negatively impacting the performance and robustness of deep learning models. Traditionally, this problem has been addressed by applying filtering algorithms to the event stream, but this may also remove some of relevant data. In this paper, we propose a novel noise-injection training methodology designed to enhance the neural networks robustness against varying levels of event noise. Our approach introduces controlled noise directly into the training data, enabling models to learn noise-resilient representations. We have conducted extensive evaluations of the proposed method using multiple benchmark datasets (N-Caltech101, N-Cars, and Mini N-ImageNet) and various network architectures, including Convolutional Neural Networks, Vision Transformers, Spiking Neural Networks, and Graph Convolutional Networks. Experimental results show that our noise-injection training strategy achieves stable performance over a range of noise intensities, consistently outperforms event-filtering techniques, and achieves the highest average classification accuracy, making it a viable alternative to traditional event-data filtering methods in an object classification system. Code: https://github.com/vision-agh/DVS_Filtering

Kamil Jeziorek, Tomasz Kryjak

Event cameras offer significant advantages over traditional frame-based sensors. These include microsecond temporal resolution, robustness under varying lighting conditions and low power consumption. Nevertheless, the effective processing of their sparse, asynchronous event streams remains challenging. Existing approaches to this problem can be categorised into two distinct groups. The first group involves the direct processing of event data with neural models, such as Spiking Neural Networks or Graph Convolutional Neural Networks. However, this approach is often accompanied by a compromise in terms of qualitative performance. The second group involves the conversion of events into dense representations with handcrafted aggregation functions, which can boost accuracy at the cost of temporal fidelity. This paper introduces a novel Self-Supervised Event Representation (SSER) method leveraging Gated Recurrent Unit (GRU) networks to achieve precise per-pixel encoding of event timestamps and polarities without temporal discretisation. The recurrent layers are trained in a self-supervised manner to maximise the fidelity of event-time encoding. The inference is performed with event representations generated asynchronously, thus ensuring compatibility with high-throughput sensors. The experimental validation demonstrates that SSER outperforms aggregation-based baselines, achieving improvements of 2.4% mAP and 0.6% on the Gen1 and 1 Mpx object detection datasets. Furthermore, the paper presents the first hardware implementation of recurrent representation for event data on a System-on-Chip FPGA, achieving sub-microsecond latency and power consumption between 1-2 W, suitable for real-time, power-efficient applications. Code is available at https://github.com/vision-agh/RecRepEvent.

Kamil Jeziorek, Piotr Wzorek, Krzysztof Blachut et al.

The utilisation of event cameras represents an important and swiftly evolving trend aimed at addressing the constraints of traditional video systems. Particularly within the automotive domain, these cameras find significant relevance for their integration into embedded real-time systems due to lower latency and energy consumption. One effective approach to ensure the necessary throughput and latency for event processing is through the utilisation of graph convolutional networks (GCNs). In this study, we introduce a custom EFGCN (Event-based FPGA-accelerated Graph Convolutional Network) designed with a series of hardware-aware optimisations tailored for PointNetConv, a graph convolution designed for point cloud processing. The proposed techniques result in up to 100-fold reduction in model size compared to Asynchronous Event-based GNN (AEGNN), one of the most recent works in the field, with a relatively small decrease in accuracy (2.9% for the N-Caltech101 classification task, 2.2% for the N-Cars classification task), thus following the TinyML trend. We implemented EFGCN on a ZCU104 SoC FPGA platform without any external memory resources, achieving a throughput of 13.3 million events per second (MEPS) and real-time partially asynchronous processing with low latency. Our approach achieves state-of-the-art performance across multiple event-based classification benchmarks while remaining highly scalable, customisable and resource-efficient. We publish both software and hardware source code in an open repository: https://github.com/vision-agh/gcnn-dvs-fpga

Konrad Lis, Tomasz Kryjak, Marek Gorgon

This paper presents LiFT, a lightweight, fully quantized 3D object detection algorithm for LiDAR data, optimized for real-time inference on FPGA platforms. Through an in-depth analysis of FPGA-specific limitations, we identify a set of FPGA-induced constraints that shape the algorithm's design. These include a computational complexity limit of 30 GMACs (billion multiply-accumulate operations), INT8 quantization for weights and activations, 2D cell-based processing instead of 3D voxels, and minimal use of skip connections. To meet these constraints while maximizing performance, LiFT combines novel mechanisms with state-of-the-art techniques such as reparameterizable convolutions and fully sparse architecture. Key innovations include the Dual-bound Pillar Feature Net, which boosts performance without increasing complexity, and an efficient scheme for INT8 quantization of input features. With a computational cost of just 20.73 GMACs, LiFT stands out as one of the few algorithms targeting minimal-complexity 3D object detection. Among comparable methods, LiFT ranks first, achieving an mAP of 51.84% and an NDS of 61.01% on the challenging NuScenes validation dataset. The code will be available at https://github.com/vision-agh/lift.

Kamil Jeziorek, Piotr Wzorek, Krzysztof Blachut et al.

Event-based vision is an emerging research field involving processing data generated by Dynamic Vision Sensors (neuromorphic cameras). One of the latest proposals in this area are Graph Convolutional Networks (GCNs), which allow to process events in its original sparse form while maintaining high detection and classification performance. In this paper, we present the hardware implementation of a~graph generation process from an event camera data stream, taking into account both the advantages and limitations of FPGAs. We propose various ways to simplify the graph representation and use scaling and quantisation of values. We consider both undirected and directed graphs that enable the use of PointNet convolution. The results obtained show that by appropriately modifying the graph representation, it is possible to create a~hardware module for graph generation. Moreover, the proposed modifications have no significant impact on object detection performance, only 0.08% mAP less for the base model and the N-Caltech data set.Finally, we describe the proposed hardware architecture of the graph generation module.

Marcin Kowlaczyk, Tomasz Kryjak

The field of neuromorphic vision is developing rapidly, and event cameras are finding their way into more and more applications. However, the data stream from these sensors is characterised by significant noise. In this paper, we propose a method for event data that is capable of removing approximately 99\% of noise while preserving the majority of the valid signal. We have proposed four algorithms based on the matrix of infinite impulse response (IIR) filters method. We compared them on several event datasets that were further modified by adding artificially generated noise and noise recorded with dynamic vision sensor. The proposed methods use about 30KB of memory for a sensor with a resolution of 1280 x 720 and is therefore well suited for implementation in embedded devices.

Michal Danilowicz, Tomasz Kryjak

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.

Hiroshi Nakano, Krzysztof Blachut, Kamil Jeziorek et al.

As the quantities of data recorded by embedded edge sensors grow, so too does the need for intelligent local processing. Such data often comes in the form of time-series signals, based on which real-time predictions can be made locally using an AI model. However, a hardware-software approach capable of making low-latency predictions with low power consumption is required. In this paper, we present a hardware implementation of an event-graph neural network for time-series classification. We leverage an artificial cochlea model to convert the input time-series signals into a sparse event-data format that allows the event-graph to drastically reduce the number of calculations relative to other AI methods. We implemented the design on a SoC FPGA and applied it to the real-time processing of the Spiking Heidelberg Digits (SHD) dataset to benchmark our approach against competitive solutions. Our method achieves a floating-point accuracy of 92.7% on the SHD dataset for the base model, which is only 2.4% and 2% less than the state-of-the-art models with over 10% and 67% fewer model parameters, respectively. It also outperforms FPGA-based spiking neural network implementations by 19.3% and 4.5%, achieving 92.3% accuracy for the quantised model while using fewer computational resources and reducing latency.

Hubert Szolc, Karol Desnos, Tomasz Kryjak

Deep reinforcement learning (DRL) is currently the most popular AI-based approach to autonomous vehicle control. An agent, trained for this purpose in simulation, can interact with the real environment with a human-level performance. Despite very good results in terms of selected metrics, this approach has some significant drawbacks: high computational requirements and low explainability. Because of that, a DRL-based agent cannot be used in some control tasks, especially when safety is the key issue. Therefore we propose to use Tangled Program Graphs (TPGs) as an alternative for deep reinforcement learning in control-related tasks. In this approach, input signals are processed by simple programs that are combined in a graph structure. As a result, TPGs are less computationally demanding and their actions can be explained based on the graph structure. In this paper, we present our studies on the use of TPGs as an alternative for DRL in control-related tasks. In particular, we consider the problem of navigating an unmanned aerial vehicle (UAV) through the unknown environment based solely on the on-board LiDAR sensor. The results of our work show promising prospects for the use of TPGs in control related-tasks.

Piotr Wzorek, Kamil Jeziorek, Tomasz Kryjak et al.

Event cameras are becoming increasingly popular as an alternative to traditional frame-based vision sensors, especially in mobile robotics. Taking full advantage of their high temporal resolution, high dynamic range, low power consumption and sparsity of event data, which only reflects changes in the observed scene, requires both an efficient algorithm and a specialised hardware platform. A recent trend involves using Graph Convolutional Neural Networks (GCNNs) implemented on a heterogeneous SoC FPGA. In this paper we focus on optimising hardware modules for graph convolution to allow flexible selection of the FPGA resource (BlockRAM, DSP and LUT) for their implementation. We propose a ''two-step convolution'' approach that utilises additional BRAM buffers in order to reduce up to 94% of LUT usage for multiplications. This method significantly improves the scalability of GCNNs, enabling the deployment of models with more layers, larger graphs sizes and their application for more dynamic scenarios.

Mateusz Wasala, Mateusz Smolarczyk, Michal Danilowicz et al.

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 x 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.

Hubert Szolc, Mateusz Wasala, Remigiusz Mietla et al.

The use of unmanned aerial vehicles (UAVs) for smart agriculture is becoming increasingly popular. This is evidenced by recent scientific works, as well as the various competitions organised on this topic. Therefore, in this work we present a system for automatic fruit counting using UAVs. To detect them, our solution uses a vision algorithm that processes streams from an RGB camera and a depth sensor using classical image operations. Our system also allows the planning and execution of flight trajectories, taking into account the minimisation of flight time and distance covered. We tested the proposed solution in simulation and obtained an average score of 87.27/100 points from a total of 500 missions. We also submitted it to the UAV Competition organised as part of the ICUAS 2024 conference, where we achieved an average score of 84.83/100 points, placing 6th in a field of 23 teams and advancing to the finals.

Mateusz Wasala, Krzysztof Blachut, Hubert Szolc et al.

Nowadays, the increasing demand for maintaining high cleanliness standards in public spaces results in the search for innovative solutions. The deployment of CCTV systems equipped with modern cameras and software enables not only real-time monitoring of the cleanliness status but also automatic detection of impurities and optimisation of cleaning schedules. The Digital Twin technology allows for the creation of a virtual model of the space, facilitating the simulation, training, and testing of cleanliness management strategies before implementation in the real world. In this paper, we present the utilisation of advanced vision surveillance systems and the Digital Twin technology in cleanliness management, using a railway station as an example. The Digital Twin was created based on an actual 3D model in the Nvidia Omniverse Isaac Sim simulator. A litter detector, bin occupancy level detector, stain segmentation, and a human detector (including the cleaning crew) along with their movement analysis were implemented. A preliminary assessment was conducted, and potential modifications for further enhancement and future development of the system were identified.

Piotr Wzorek, Tomasz Kryjak

Despite the dynamic development of computer vision algorithms, the implementation of perception and control systems for autonomous vehicles such as drones and self-driving cars still poses many challenges. A video stream captured by traditional cameras is often prone to problems such as motion blur or degraded image quality due to challenging lighting conditions. In addition, the frame rate - typically 30 or 60 frames per second - can be a limiting factor in certain scenarios. Event cameras (DVS -- Dynamic Vision Sensor) are a potentially interesting technology to address the above mentioned problems. In this paper, we compare two methods of processing event data by means of deep learning for the task of pedestrian detection. We used a representation in the form of video frames, convolutional neural networks and asynchronous sparse convolutional neural networks. The results obtained illustrate the potential of event cameras and allow the evaluation of the accuracy and efficiency of the methods used for high-resolution (1280 x 720 pixels) footage.

Piotr Wzorek, Tomasz Kryjak

This paper presents a method for automatic generation of a training dataset for a deep convolutional neural network used for playing card detection. The solution allows to skip the time-consuming processes of manual image collecting and labelling recognised objects. The YOLOv4 network trained on the generated dataset achieved an efficiency of 99.8% in the cards detection task. The proposed method is a part of a project that aims to automate the process of broadcasting duplicate bridge competitions using a vision system and neural networks.

Marcin Kowalczyk, Tomasz Kryjak

This work describes the hardware implementation of a connected component labelling (CCL) module in reprogammable logic. The main novelty of the design is the "full", i.e. without any simplifications, support of a 4 pixel per clock format (4 ppc) and real-time processing of a 4K/UltraHD video stream (3840 x 2160 pixels) at 60 frames per second. To achieve this, a special labelling method was designed and a functionality that stops the input data stream in order to process pixel groups which require writing more than one merger into the equivalence table. The proposed module was verified in simulation and in hardware on the Xilinx Zynq Ultrascale+ MPSoC chip on the ZCU104 evaluation board.

Artur Cyba, Hubert Szolc, Tomasz Kryjak

In this paper, we present a control system that allows a drone to fly autonomously through a series of gates marked with ArUco tags. A simple and low-cost DJI Tello EDU quad-rotor platform was used. Based on the API provided by the manufacturer, we have created a Python application that enables the communication with the drone over WiFi, realises drone positioning based on visual feedback, and generates control. Two control strategies were proposed, compared, and critically analysed. In addition, the accuracy of the positioning method used was measured. The application was evaluated on a laptop computer (about 40 fps) and a Nvidia Jetson TX2 embedded GPU platform (about 25 fps). We provide the developed code on GitHub.

Dominika Przewlocka-Rus, Marcin Kowalczyk, Tomasz Kryjak

Deep learning algorithms are a key component of many state-of-the-art vision systems, especially as Convolutional Neural Networks (CNN) outperform most solutions in the sense of accuracy. To apply such algorithms in real-time applications, one has to address the challenges of memory and computational complexity. To deal with the first issue, we use networks with reduced precision, specifically a binary neural network (also known as XNOR). To satisfy the computational requirements, we propose to use highly parallel and low-power FPGA devices. In this work, we explore the possibility of accelerating XNOR networks for traffic sign classification. The trained binary networks are implemented on the ZCU 104 development board, equipped with a Zynq UltraScale+ MPSoC device using two different approaches. Firstly, we propose a custom HDL accelerator for XNOR networks, which enables the inference with almost 450 fps. Even better results are obtained with the second method - the Xilinx FINN accelerator - enabling to process input images with around 550 frame rate. Both approaches provide over 96% accuracy on the test set.

Joanna Stanisz, Konrad Lis, Tomasz Kryjak et al.

In this paper we present our research on the optimisation of a deep neural network for 3D object detection in a point cloud. Techniques like quantisation and pruning available in the Brevitas and PyTorch tools were used. We performed the experiments for the PointPillars network, which offers a reasonable compromise between detection accuracy and calculation complexity. The aim of this work was to propose a variant of the network which we will ultimately implement in an FPGA device. This will allow for real-time LiDAR data processing with low energy consumption. The obtained results indicate that even a significant quantisation from 32-bit floating point to 2-bit integer in the main part of the algorithm, results in 5%-9% decrease of the detection accuracy, while allowing for almost a 16-fold reduction in size of the model.

Dominika Przewlocka, Mateusz Wasala, Hubert Szolc et al.

In this paper the research on optimisation of visual object tracking using a Siamese neural network for embedded vision systems is presented. It was assumed that the solution shall operate in real-time, preferably for a high resolution video stream, with the lowest possible energy consumption. To meet these requirements, techniques such as the reduction of computational precision and pruning were considered. Brevitas, a tool dedicated for optimisation and quantisation of neural networks for FPGA implementation, was used. A number of training scenarios were tested with varying levels of optimisations - from integer uniform quantisation with 16 bits to ternary and binary networks. Next, the influence of these optimisations on the tracking performance was evaluated. It was possible to reduce the size of the convolutional filters up to 10 times in relation to the original network. The obtained results indicate that using quantisation can significantly reduce the memory and computational complexity of the proposed network while still enabling precise tracking, thus allow to use it in embedded vision systems. Moreover, quantisation of weights positively affects the network training by decreasing overfitting.

Krzysztof Blachut, Hubert Szolc, Mateusz Wasala et al.

In this paper we present a vision based hardware-software control system enabling autonomous landing of a multirotor unmanned aerial vehicle (UAV). It allows the detection of a marked landing pad in real-time for a 1280 x 720 @ 60 fps video stream. In addition, a LiDAR sensor is used to measure the altitude above ground. A heterogeneous Zynq SoC device is used as the computing platform. The solution was tested on a number of sequences and the landing pad was detected with 96% accuracy. This research shows that a reprogrammable heterogeneous computing system is a good solution for UAVs because it enables real-time data stream processing with relatively low energy consumption.

Piotr Janus, Tomasz Kryjak, Marek Gorgon

This paper presents a GPU implementation of two foreground object segmentation algorithms: Gaussian Mixture Model (GMM) and Pixel Based Adaptive Segmenter (PBAS) modified for RGB-D data support. The simultaneous use of colour (RGB) and depth (D) data allows to improve segmentation accuracy, especially in case of colour camouflage, illumination changes and occurrence of shadows. Three GPUs were used to accelerate calculations: embedded NVIDIA Jetson TX2 (Maxwell architecture), mobile NVIDIA GeForce GTX 1050m (Pascal architecture) and efficient NVIDIA RTX 2070 (Turing architecture). Segmentation accuracy comparable to previously published works was obtained. Moreover, the use of a GPU platform allowed to get real-time image processing. In addition, the system has been adapted to work with two RGB-D sensors: RealSense D415 and D435 from Intel.