ABSTRACT
Cloud-top heights (CTH) from the Multiangle Imaging Spectroradiometer (MISR) and the Moderate Resolution Imaging Spectroradiometer (MODIS) on Terra constitute our longest-running single-platform CTH record from a stable orbit. Here, we provide the first evaluation of the Terra Level 2 CTH record against collocated International Space Station Cloud-Aerosol Transport System (CATS) lidar observations between 50ºN and 50ºS. Bias and precision of Terra CTH relative to CATS is shown to be strongly tied to cloud horizontal and vertical heterogeneity and altitude. For single-layered, unbroken, optically thick clouds observed over all altitudes, the uncertainties in MODIS and MISR CTH are -540 ± 690 m and -280 ± 370 m, respectively. The uncertainties are generally smaller for lower altitude clouds and larger for optically thin clouds. For multi-layered clouds, errors are summarized herein using both absolute CTH and CATS-layer-altitude proximity to Terra CTH. We show that MISR detects the lower cloud in a two-layered system, provided top-layer optical depth <â¼0.3, but MISR low-cloud CTH errors are unaltered by the presence of thin cirrus. Systematic and random errors are propagated to explain inter-sensor disagreements, as well as to provide the first estimate of the MISR stereo-opacity bias. For MISR, altitude-dependent wind-retrieval bias (-90 to -110 m) and stereo-opacity bias (-60 to -260 m) and for MODIS, CO2-slicing bias due to geometrically thick cirrus leads to overall negative CTH bias. MISR's precision is largely driven by precision in retrieved wind-speed (3.7 m s-1), whereas MODIS precision is driven by forward-modeling uncertainty.
ABSTRACT
Adopting effective techniques to automatically detect and identify small drones is a very compelling need for a number of different stakeholders in both the public and private sectors. This work presents three different original approaches that competed in a grand challenge on the "Drone vs. Bird" detection problem. The goal is to detect one or more drones appearing at some time point in video sequences where birds and other distractor objects may be also present, together with motion in background or foreground. Algorithms should raise an alarm and provide a position estimate only when a drone is present, while not issuing alarms on birds, nor being confused by the rest of the scene. In particular, three original approaches based on different deep learning strategies are proposed and compared on a real-world dataset provided by a consortium of universities and research centers, under the 2020 edition of the Drone vs. Bird Detection Challenge. Results show that there is a range in difficulty among different test sequences, depending on the size and the shape visibility of the drone in the sequence, while sequences recorded by a moving camera and very distant drones are the most challenging ones. The performance comparison reveals that the different approaches perform somewhat complementary, in terms of correct detection rate, false alarm rate, and average precision.
Subject(s)
Deep Learning , Algorithms , Animals , Birds , MotionABSTRACT
Chest radiographs are primarily employed for the screening of pulmonary and cardio-/thoracic conditions. Being undertaken at primary healthcare centers, they require the presence of an on-premise reporting Radiologist, which is a challenge in low and middle income countries. This has inspired the development of machine learning based automation of the screening process. While recent efforts demonstrate a performance benchmark using an ensemble of deep convolutional neural networks (CNN), our systematic search over multiple standard CNN architectures identified single candidate CNN models whose classification performances were found to be at par with ensembles. Over 63 experiments spanning 400 hours, executed on a 11.3 FP32 TensorTFLOPS compute system, we found the Xception and ResNet-18 architectures to be consistent performers in identifying co-existing disease conditions with an average AUC of 0.87 across nine pathologies. We conclude on the reliability of the models by assessing their saliency maps generated using the randomized input sampling for explanation (RISE) method and qualitatively validating them against manual annotations locally sourced from an experienced Radiologist. We also draw a critical note on the limitations of the publicly available CheXpert dataset primarily on account of disparity in class distribution in training vs. testing sets, and unavailability of sufficient samples for few classes, which hampers quantitative reporting due to sample insufficiency.
Subject(s)
Lung , Neural Networks, Computer , Radiography , Reproducibility of Results , ResearchABSTRACT
Particle Swarm Optimization (PSO), a population based technique for stochastic search in a multidimensional space, has so far been employed successfully for solving a variety of optimization problems including many multifaceted problems, where other popular methods like steepest descent, gradient descent, conjugate gradient, Newton method, etc. do not give satisfactory results. Herein, we propose a modified PSO algorithm for unbiased global minima search by integrating with density functional theory which turns out to be superior to the other evolutionary methods such as simulated annealing, basin hopping and genetic algorithm. The present PSO code combines evolutionary algorithm with a variational optimization technique through interfacing of PSO with the Gaussian software, where the latter is used for single point energy calculation in each iteration step of PSO. Pure carbon and carbon containing systems have been of great interest for several decades due to their important role in the evolution of life as well as wide applications in various research fields. Our study shows how arbitrary and randomly generated small Cn clusters (n = 3-6, 10) can be transformed into the corresponding global minimum structure. The detailed results signify that the proposed technique is quite promising in finding the best global solution for small population size clusters.
