Coin grading is a crucial process in numismatics that determines a coin’s value on the basis of its condition. However, the traditional grading procedure is performed by human which makes it subjective and time-consuming, with different experts potentially assigning different grades to the same coin. This paper introduces a Deep Learning (DL) model for the type-invariant automatic coin grading across the full 70-points Sheldon scale. We propose a Siamese neural network based on EfficientNet encoders for dual-sided coin evaluation, which performs the analysis of both the obverse and reverse images of coins. The model is trained and evaluated on a comprehensive dataset containing 36,205 unique professionally graded samples covering all 30 grades from Poor (P)1 to Mint State (MS)70 across the multiple coin types. The proposed DL solution achieves 0.3403 accuracy and 1.52 Mean Absolute Error (MAE) along the Sheldon scale, outperforming both the Machine Learning (ML) baseline and the state-of-the-art approaches. Unlike earlier studies that were limited to fewer coin types or reduced grading scales, our work addresses the full Sheldon scale grading across the multiple coin types, providing a more robust and generalizable solution for automatic coin grading.

At present, there is a significant focus on improving agricultural productivity, which is essential both on a national and global scale. There are many factors that impact the quality and quantity of crop yields, and environmental changes such as weather conditions require careful consideration and quick action to avoid significant losses. When it comes to managing fruit orchards, large areas need to be monitored, and unmanned aerial vehicles (UAVs) are currently being used to automate and enhance the process. UAVs can capture images that can be analyzed using neural-based methods to extract useful information about the state of the trees in the orchard. However, there are some challenges that must be addressed, such as the scarcity of open-access labeled datasets and the imbalanced distribution of target classes, which include rare events or anomalous vegetation states. To tackle these issues, in this paper, a unique dataset of apple trees observations captured by UAVs has been collected and shared to the research community. This dataset includes healthy and unhealthy trees with formed and unformed crowns. Experiments were conducted using YOLOv5 neural network for object detection to evaluate its effectiveness in solving agricultural remote sensing tasks. To adjust model’s performance, we proposed a task-specific transfer learning approach that involves pre-training the model on a synthetic dataset. The synthetic dataset was generated using object-based augmentation (OBA) with the original target objects. Thereby, instead of utilizing pre-trained weights from the general domain such as COCO dataset, we use not only domain-specific UAV-derived data for pre-training but the images for the same task of apple tree health examining. The proposed approach allowed us to increase the mAP from 0.642 to 0.706 compared with the conventional approach.

A number of polyolefin plastics including Polypropylene (PP), Polyethylene (PE), Polystyrene (PS) are widespread plastics worldwide. At the same time, they are frequently recycled plastics. Still a considerable amount of waste plastics come in a mixed mode imposing technological challenges for recycling. Households cannot provide thorough separation on the early stage, especially if the plastic products are not marked properly. In this paper, we report on the analysis of two pair of plastic shavers of the similar color and shape while made of different plastics. For this reason an experimental testbed based on a Raman Spectrometer with an embedded microscope has been designed. The proposed solution enables the identification of PP and PS by the sensor system with the singular spectral range 1320 nm - 1740 nm. There is an opportunity to measure three spectral ranges simultaneously.

Video gaming and eSports is a quickly developing industry already involving billions of players worldwide. Gaming and eSports tournaments require strong mental abilities to avoid severe stress and other negative consequences upon completing the game. In this article, we report on the impact of emotions on a team performance. For this reason, we collect audio recordings and game logs from the players in real conditions at an eSports tournament. This data is further used in trained machine learning models for analysis of players' emotional conditions from the voice during the game. We considered recognition of several types of emotions as well as the background sounds. To do this, we trained 92.7% accuracy classifier of six most common classes of emotions and sounds in eSports audio and applied it to eSports data. As a result, we demonstrate that there is an opportunity to measure the eSports team's performance from the players' emotional conditions obtained from the voice communication. We found that there is a strong correlation among the performance of the team, communication between the players, and emotional sentiment of communication. The teams achieve much better results when they had much more internal conversations during the game.

Wearable devices continuously gain their popularity due to flexibility and opportunity to enable more applications and services, in particular, in healthcare industry. Guarantying long-term operation for these devices is among the key challenges at the design stage. In this work, we propose a multi-source energy harvesting system to address this problem. The proposed solution is based on two technologies namely an Enzymatic Biofuel Cell (BFC) and a low-temperature gradient Thermoelectric Generator (TEG) which are capable of generating power from human perspiration and heat, respectively. Upon designing BFC, we designed a functional circuit that combines it with the TEG. The resulting BFC has an output voltage of 200 mV and a power of about 7 nW. An array of BFCs is used as a source for initial push to start up the internal logical control circuit of the DC-DC controller. The low-temperature gradient TEG with an output voltage ranging from 40 mV to 70 mV and a constant measured output power of 160 uW is used together with the BFCs for energy harvesting upon the initial starting of the DC-DC controller.

In most cases, being an ill-posed problem, image restoration opts to restore a high-quality image from a low-quality one, assuming that some degradation model produced given low-quality input. A lot of restoration methods were proposed for the case when linear degradation operator and i.i.d. Gaussian likelihood is assumed. However, such methods are known not to generalize well. They show a sub-par performance on real data, for which the actual degradation model is neither linear nor even exactly known. The state-of-the-art machine learning allows for overcoming this issue and learn a restoration model to the real data. The main drawback of such approaches is overfitting since to learn an inverse mapping between the low-quality and high-quality samples, they rely entirely on data. They do not utilize limited but existing knowledge of how degradation was performed. In this paper, we study learned gradient descent based image restoration and synthesis. Both linear and non-linear known restoration problems are considered, and research on how a known degradation model may be incorporated in a learned gradient-based restoration procedure is provided. Our results demonstrate that explicit usage of the degradation model and its learned linear and non-linear approximations boost restoration quality compared to a baseline without this feature.

Artificial Intelligence (AI) is a widely used tool in precision agriculture for estimating the quality of food. It is especially relevant while assessing crops at various harvest and postharvest stages. Crop disease and damage detection is a task of top priority: some postharvest diseases or damages, e.g. decay, may destroy the crops and create toxins harmful to human beings. In this work, we apply U-Net, Deeplab, and Mask R-CNN models based on Convolutional Neural Networks (CNNs) for detecting and predicting the postharvest decay areas in stored apple fruits. Novelty of our approach is separate segmentation and prediction of postharvest decay and non-decay areas in apples. Images were acquired with a custom-made testbed consisting of a digital camera, stepper drivers pallet for apples, and the PC. The dataset of the acquired time-sequenced images of the postharvest apples includes 4440 images and is available online. Mask R-CNN demonstrated the best performance and achieved 98.81% of the mean Average Precision (mAP) for apples and 43.60% of the mAP for postharvest decay zones. The proposed approach is promising for improving the food storage process in precision agriculture.

Current research in eSports lacks the tools for proper game practising and performance analytics. The majority of prior work relied only on in-game data for advising the players on how to perform better. However, in-game mechanics and trends are frequently changed by new patches limiting the lifespan of the models trained exclusively on the in-game logs. In this article, we propose the methods based on the sensor data analysis for predicting whether a player will win the future encounter. The sensor data were collected from ten participants in 22 matches in the League of Legends video game. We have trained machine learning models, including the transformer and gated recurrent unit, to predict whether the player wins the encounter taking place after some fixed time in the future. For 10-s forecasting horizon, the transformer neural network architecture achieves the ROC AUC score of 0.706. This model is further developed into the detector capable of predicting that a player will lose the encounter occurring in 10 s in 88.3% of cases with 73.5% accuracy. This might be used as a players’ burnout or fatigue detector, advising players to retreat. We have also investigated which physiological features affect the chance to win or lose the next in-game encounter.

The Hogweed of Sosnowskyi (lat. Heracleum sosnówskyi) is poisonous for humans, dangerous for farming crops, and local ecosystems. This plant is fast-growing and has already spread all over Eurasia: from Germany to the Siberian part of Russia, and its distribution expands year-by-year. In-situ detection of this harmful plant is a tremendous challenge for many countries. Meanwhile, there are no automatic systems for detection and localization of hogweed. In this article, we report on an approach for fast and accurate detection of hogweed. The approach includes the Unmanned Aerial Vehicle (UAV) with an embedded system on board running various Fully Convolutional Neural Networks (FCNN). We propose the optimal architecture of FCNN for the embedded system relying on the trade-off between the detection quality and frame rate. We propose a model that achieves ROC AUC 0.96 in the hogweed segmentation task, which can process 4K frames at 0.46 FPS on NVIDIA Jetson Nano. The developed system can recognize the hogweed on the scale of individual plants and leaves. This system opens up a wide vista for obtaining comprehensive and relevant data about the spreading of harmful plants allowing for the elimination of their expansion.

Unmanned Aerial Vehicles (UAV) have been recently applied in several various civilian applications. Based on this, there is a growing need for intuitive UAV control interfaces. In this work, we report on the Body-Machine Interface (BMI), helping a human operator to control a quadrotor through the gesture commands. We perform the human motion capture through wearable sensors and Kalman filter to reduce the noise. For the gesture command recognition, we designed the Recurrent Neural Network recognizing gestures within 65 ms. For the quadrotor orientation estimation, we designed the Extended Kalman Filter (EKF). We assess the proposed BMI via the simulations and experiments: the standard deviation of the trajectories varies for up to 10 cm.

Aerial drones can be used for a number of monitoring and control applications. Most of existing drone control platforms are quite primitive in terms of body-machine interface. They are usually a variation of a hand-held remote controller or ground control station. However, in a number of line-of-sight scenarios it would be more convenient to use the human gestures and voice for the drone control. In this work, we present an approach for instantaneous control of drones based on human voice and gestures. The proposed solution includes wearable sensors and embedded artificial intelligence. We use a microphone and an Inertial Measurement Unit (IMU) for capturing the human voice and the hand movements. Primary control is implemented by a voice recognition unit based on Recurrent Neural Network (RNN) while the secondary control is implemented by the gesture recognition system based on Convolutional Neural Network (CNN). For implementing the embedded intelligence, we use a low-power embedded system with a graphical processing unit able to run pre-trained neural networks on board of the drone. As a result, the system can perform different speech and gesture recognition tasks real-time.

eSports industry has greatly progressed within the last decade in terms of audience and fund rising, broadcasting, networking and hardware. Since the number and quality of professional team has evolved too, there is a reasonable need in improving skills and training process of professional eSports athletes. In this work, we demonstrate a system able to collect heterogeneous data (physiological, environmental, video, telemetry) and guarantying synchronization with 10 ms accuracy. In particular, we demonstrate how to synchronize various sensors and ensure post synchronization, i.e. logged video, a so-called demo file, with the sensors data. Our experimental results achieved on the CS:GO game discipline show up to 3 ms accuracy of the time synchronization of the gaming computer.

We report on the teleoperation system DronePick which provides remote object picking and delivery by a human- controlled quadcopter. The main novelty of the proposed system is that the human user continuously gets the visual and haptic feedback for accurate teleoperation. DronePick consists of a quadcopter equipped with a magnetic grabber, a tactile glove with finger motion tracking sensor, hand tracking system, and the Virtual Reality (VR) application. The human operator teleoperates the quadcopter by changing the position of the hand. The proposed vibrotactile patterns representing the location of the remote object relative to the quadcopter are delivered to the glove. It helps the operator to determine when the quadcopter is right above the object. When the “pick” command is sent by clasping the hand in the glove, the quadcopter decreases its altitude and the magnetic grabber attaches the target object. The whole scenario is in parallel simulated in VR. The air flow from the quadcopter and the relative positions of VR objects help the operator to determine the exact position of the delivered object to be picked. The experiments showed that the vibrotactile patterns were recognized by the users at the high recognition rates: the average 99% recognition rate and the average 2.36s recognition time. The real-life implementation of DronePick featuring object picking and delivering to the human was developed and tested.

In this work, we present the design and prototype of a selfpowered sensor node for the aircraft structural health monitoring. The sensor node is powered by the ambient vibrations generated by the aircraft wings. Sensing devices perform the comprehensive condition monitoring of structures and systems, as well as measurement of environmental parameters, e.g. ambient temperature, ionizing radiation levels, to help operators in assessing the aircraft status at every stage of its mission. With the wireless communication the sensor nodes can be networked with no need for wiring which implies extra weight on board.

Wireless sensor networks (WSNs) have recently been applied for industrial monitoring, including combustible and flammable gases monitoring. In this work, we present a wireless gas sensor node in which a widely used Wheatstone sensing circuit based on two sensors is exchanged with a single sensor circuit, as well as the associate gas measurement procedure. The core of the measurement procedure is the four-stage heating profile, which enables low power consumption of sensing circuit and thermo compensation adjustment. A thermo compensation algorithm is capable of avoiding the effect of the environmental temperature on the measurements by keeping stable zero-offset within ±1 mV and ensuring low absolute error within 0.1% vol. The thorough design of the sensor node allows it to fit into the 5.5 cm3packaging, which ensures its true ubiquitous deployment in outdoor and industrial environment.

Energy consumption is one of the most constraining requirements for the design and implementation of wireless sensor networks. Simulation tools allow one to significantly decrease the effort and time spent to choose the right solution. Existing simulators provide varying degrees of analysis for communication, application and energy domains. However, they do not provide enough flexibility to estimate the consumed power for a wide range of wireless sensor network (WSN) hardware (HW) platforms. In this paper we present a flexible and extensible simulation framework to estimate power consumption of sensor network applications for arbitrary HW platforms. This framework allows designers of sensor networks to estimate power consumption of the explored HW platform which permits the selection of an optimal HW solution and software (SW) implementation for the desired projects.