Litter Detection with Deep Learning: A Comparative Study.

Pollution in the form of litter in the natural environment is one of the great challenges of our times. Automated litter detection can help assess waste occurrences in the environment. Different machine learning solutions have been explored to develop litter detection tools, thereby supporting research, citizen science, and volunteer clean-up initiatives. However, to the best of our knowledge, no work has investigated the performance of state-of-the-art deep learning object detection approaches in the context of litter detection. In particular, no studies have focused on the assessment of those methods aiming their use in devices with low processing capabilities, e.g., mobile phones, typically employed in citizen science activities. In this paper, we fill this literature gap. We performed a comparative study involving state-of-the-art CNN architectures (e.g., Faster RCNN, Mask-RCNN, EfficientDet, RetinaNet and YOLO-v5), two litter image datasets and a smartphone. We also introduce a new dataset for litter detection, named PlastOPol, composed of 2418 images and 5300 annotations. The experimental results demonstrate that object detectors based on the YOLO family are promising for the construction of litter detection solutions, with superior performance in terms of detection accuracy, processing time, and memory footprint.

Cognitive Control-Loop for Elastic Optical Networks with Space-Division Multiplexing.

This paper introduces a new solution to improve network performance by decreasing spectrum fragmentation, crosstalk interference, blocking of virtual networks, cost, and link load imbalance. These problems degrade the performance of Elastic Optical Networks with Space-Division Multiplexing. The proposed solution, called Cognitive control loop (CO-OP), is capable of identifying a set of problems and creating plans to mitigate these problems. The CO-OP comprises four functions that employ learning algorithms to identify problems and plan a series of actions to reduce or eliminate them. The results show that the CO-OP can effectively decrease up to 30% the blocking of requests and up to 50% the crosstalk occurrence compared to existing algorithms.

Automatic Markerless Motion Detector Method against Traditional Digitisation for 3-Dimensional Movement Kinematic Analysis of Ball Kicking in Soccer Field Context.

Kicking is a fundamental skill in soccer that often contributes to match outcomes. Lower limb movement features (e.g., joint position and velocity) are determinants of kick performance. However, obtaining kicking kinematics under field conditions generally requires time-consuming manual tracking. The current study aimed to compare a contemporary markerless automatic motion estimation algorithm (OpenPose) with manual digitisation (DVIDEOW software) in obtaining on-field kicking kinematic parameters. An experimental dataset of under-17 players from all outfield positions was used. Kick attempts were performed in an official pitch against a goalkeeper. Four digital video cameras were used to record full-body motion during support and ball contact phases of each kick. Three-dimensional positions of hip, knee, ankle, toe and foot centre-of-mass (CMfoot) generally showed no significant differences when computed by automatic as compared to manual tracking (whole kicking movement cycle), while only z-coordinates of knee and calcaneus markers at specific points differed between methods. The resulting time-series matrices of positions (r2 = 0.94) and velocity signals (r2 = 0.68) were largely associated (all p < 0.01). The mean absolute error of OpenPose motion tracking was 3.49 cm for determining positions (ranging from 2.78 cm (CMfoot) to 4.13 cm (dominant hip)) and 1.29 m/s for calculating joint velocity (0.95 m/s (knee) to 1.50 m/s (non-dominant hip)) as compared to reference measures by manual digitisation. Angular range-of-motion showed significant correlations between methods for the ankle (r = 0.59, p < 0.01, large) and knee joint displacements (r = 0.84, p < 0.001, very large) but not in the hip (r = 0.04, p = 0.85, unclear). Markerless motion tracking (OpenPose) can help to successfully obtain some lower limb position, velocity, and joint angular outputs during kicks performed in a naturally occurring environment.

Detecting face presentation attacks in mobile devices with a patch-based CNN and a sensor-aware loss function.

With the widespread use of biometric authentication comes the exploitation of presentation attacks, possibly undermining the effectiveness of these technologies in real-world setups. One example takes place when an impostor, aiming at unlocking someone else's smartphone, deceives the built-in face recognition system by presenting a printed image of the user. In this work, we study the problem of automatically detecting presentation attacks against face authentication methods, considering the use-case of fast device unlocking and hardware constraints of mobile devices. To enrich the understanding of how a purely software-based method can be used to tackle the problem, we present a solely data-driven approach trained with multi-resolution patches and a multi-objective loss function crafted specifically to the problem. We provide a careful analysis that considers several user-disjoint and cross-factor protocols, highlighting some of the problems with current datasets and approaches. Such analysis, besides demonstrating the competitive results yielded by the proposed method, provides a better conceptual understanding of the problem. To further enhance efficacy and discriminability, we propose a method that leverages the available gallery of user data in the device and adapts the method decision-making process to the user's and the device's own characteristics. Finally, we introduce a new presentation-attack dataset tailored to the mobile-device setup, with real-world variations in lighting, including outdoors and low-light sessions, in contrast to existing public datasets.

RadialPheno: A tool for near-surface phenology analysis through radial layouts.

PREMISE: Increasingly, researchers studying plant phenology are exploring novel technologies to remotely observe plant changes over time. The increasing use of phenocams to monitor leaf phenology, based on the analysis of indices extracted from sequences of daily digital vegetation images, has demanded the development of appropriate tools for data visualization and analysis. Here, we describe RadialPheno, a tool that uses radial layouts to represent time series from digital repeat photographs, and applies them to the analysis of leafing patterns and leaf exchange strategies of different vegetations. METHODS AND RESULTS: We developed a web tool, RadialPheno, provided with the R and Shiny environments, which uses radial visual structures to represent cyclical multidimensional temporal data associated with digital image time series. We demonstrate the application of our methods and tool for a savanna vegetation phenology in the Brazilian Cerrado. We visually represented the greenness index extracted from sequential imagery using the RadialPheno tool. CONCLUSIONS: RadialPheno was successfully applied for the visualization and interpretation of individual, species, and community long-term leafing phenology data associated with near-surface phenological observations of Cerrado vegetation. RadialPheno was also effective for intercomparisons of ground-based direct visual observations and camera-derived phenology observations.

Pixel-Level Tissue Classification for Ultrasound Images.

BACKGROUND: Pixel-level tissue classification for ultrasound images, commonly applied to carotid images, is usually based on defining thresholds for the isolated pixel values. Ranges of pixel values are defined for the classification of each tissue. The classification of pixels is then used to determine the carotid plaque composition and, consequently, to determine the risk of diseases (e.g., strokes) and whether or not a surgery is necessary. The use of threshold-based methods dates from the early 2000s but it is still widely used for virtual histology. METHODOLOGY/PRINCIPAL FINDINGS: We propose the use of descriptors that take into account information about a neighborhood of a pixel when classifying it. We evaluated experimentally different descriptors (statistical moments, texture-based, gradient-based, local binary patterns, etc.) on a dataset of five types of tissues: blood, lipids, muscle, fibrous, and calcium. The pipeline of the proposed classification method is based on image normalization, multiscale feature extraction, including the proposal of a new descriptor, and machine learning classification. We have also analyzed the correlation between the proposed pixel classification method in the ultrasound images and the real histology with the aid of medical specialists. CONCLUSIONS/SIGNIFICANCE: The classification accuracy obtained by the proposed method with the novel descriptor in the ultrasound tissue images (around 73%) is significantly above the accuracy of the state-of-the-art threshold-based methods (around 54%). The results are validated by statistical tests. The correlation between the virtual and real histology confirms the quality of the proposed approach showing it is a robust ally for the virtual histology in ultrasound images.

Mid-level image representations for real-time heart view plane classification of echocardiograms.

In this paper, we explore mid-level image representations for real-time heart view plane classification of 2D echocardiogram ultrasound images. The proposed representations rely on bags of visual words, successfully used by the computer vision community in visual recognition problems. An important element of the proposed representations is the image sampling with large regions, drastically reducing the execution time of the image characterization procedure. Throughout an extensive set of experiments, we evaluate the proposed approach against different image descriptors for classifying four heart view planes. The results show that our approach is effective and efficient for the target problem, making it suitable for use in real-time setups. The proposed representations are also robust to different image transformations, e.g., downsampling, noise filtering, and different machine learning classifiers, keeping classification accuracy above 90%. Feature extraction can be performed in 30 fps or 60 fps in some cases. This paper also includes an in-depth review of the literature in the area of automatic echocardiogram view classification giving the reader a through comprehension of this field of study.