Bounding overwatch

About: Bounding overwatch is a research topic. Over the lifetime, 966 publications have been published within this topic receiving 15156 citations.

Revisiting Gini for equitable humanitarian logistics

Abstract: Modeling equity in the allocation of scarce resources is a fast-growing concern in the humanitarian logistics field. The Gini coefficient is one of the most widely recognized measures of inequity and it was originally characterized by means of the Lorenz curve, which is a mathematical function that links the cumulative share of income to rank-ordered groups in a population. So far, humanitarian logistics models that have approached equity using the Gini coefficient do not actually optimize its original formulation, but they use alternative definitions that do not necessarily replicate that original Gini measure. In this paper, we derive the original Gini coefficient via the Lorenz curve to optimize the effectiveness-equity trade-off in a humanitarian location-allocation problem. We also propose new valid inequalities based on a bounding Lorenz curve to tighten the linear relaxation of our model and develop a clustering-based construction of the Lorenz curve that requires fewer additional constraints and variables than the original one. The computational study, based on the floods and landslides in Rio de Janeiro state, Brazil, reveals that while alternative Gini definitions have interesting properties, they can generate vastly different decisions compared to the original Gini coefficient. In addition, viewed from the perspective of the original Gini coefficient, these decisions can be significantly less equitable.

Accelerated Point-Wise Maximum Approach to Approximate Dynamic Programming

Abstract: In this article, we describe an approximate dynamic programming (ADP) approach to compute lower bounds on the optimal value function for a discrete time, continuous space, and infinite horizon setting. The approach iteratively constructs a family of lower bounding approximate value functions by using the so-called Bellman inequality. The novelty of our approach is that, at each iteration, we aim to compute an approximate value function that maximizes the point-wise maximum taken with the family of approximate value functions computed thus far. This leads to a nonconvex objective, and we propose a gradient ascent algorithm to find stationary points by solving a sequence of convex optimization problems. We provide convergence guarantees for our algorithm and an interpretation for how the gradient computation relates to the state-relevance weighting parameter appearing in related ADP approaches. We demonstrate through numerical examples that, when compared to the existing approaches, the algorithm we propose computes tighter suboptimality bounds with comparable computation time.

Multi-to-multi tracking in video analytics

Abstract: Techniques and systems are provided for processing video data. For example, techniques and systems are provided for matching a plurality of bounding boxes to a plurality of trackers. In some examples, a first association is performed, in which case one or more of the plurality of bounding boxes are associated with one or more of the plurality of trackers by minimizing distances between the one or more bounding boxes and the one or more trackers. A set of unmatched trackers are identified from the plurality of trackers after the first association. The set of unmatched trackers are not associated with a bounding box from the plurality of bounding boxes during the first association. A second association is then performed, in which case each of the set of unmatched trackers is associated with an associated bounding box from the plurality of bounding boxes that is within a first pre-determined distance. A set of unmatched bounding boxes is identified from the plurality of bounding boxes after the second association. The set of unmatched bounding boxes are not associated with a tracker from the plurality of trackers during the second association. A third association is then performed, in which case each of the set of unmatched bounding boxes is associated with an associated tracker from the plurality of trackers that is within a second pre-determined distance.

JRDB-Act: A Large-scale Dataset for Spatio-temporal Action, Social Group and Activity Detection

Abstract: The availability of large-scale video action understanding datasets has facilitated advances in the interpretation of visual scenes containing people. However, learning to recognise human actions and their social interactions in an unconstrained real-world environment comprising numerous people, with potentially highly unbalanced and long-tailed distributed action labels from a stream of sensory data captured from a mobile robot platform remains a significant challenge, not least owing to the lack of a reflective large-scale dataset. In this paper, we introduce JRDB-Act, as an extension of the existing JRDB, which is captured by a social mobile manipulator and reflects a real distribution of human daily-life actions in a university campus environment. JRDB-Act has been densely annotated with atomic actions, comprises over 2.8M action labels, constituting a large-scale spatio-temporal action detection dataset. Each human bounding box is labeled with one pose-based action label and multiple~(optional) interaction-based action labels. Moreover JRDB-Act provides social group annotation, conducive to the task of grouping individuals based on their interactions in the scene to infer their social activities~(common activities in each social group). Each annotated label in JRDB-Act is tagged with the annotators' confidence level which contributes to the development of reliable evaluation strategies. In order to demonstrate how one can effectively utilise such annotations, we develop an end-to-end trainable pipeline to learn and infer these tasks, i.e. individual action and social group detection. The data and the evaluation code is publicly available at this https URL.

DeepTrac: Applying Artificial Intelligence in Plant Disease Detection

Abstract: Small community local farmers face many unique issues in farming such as irrigating uneven farming lands, measuring soil water level, plant health monitoring, everyday trips to inspect for weed and pests, high cost and low-quality products, and many more. Most critical and challenging problem in farming is the early detection of plant diseases. In this research, an AI-Deep learning based automated plant disease detection method using FPN with Faster R-CNN architecture is proposed. In this work, we evaluated our proposed method on disease detection with main outcome measure of Intersection over Union (IoU) ratio (ratio of overlap between predicted and annotated bounding boxes), precision (ratio of true predictions to total predictions), recall (ratio of true predictions to annotated bounding boxes) and disease spread direction. Our method is experimented on Bacterial Spot detection in bell-pepper plant leaf images obtained from the PlantVillage dataset. The proposed model achieved average precision 100% and average recall 99.7% for an IoU ratio of 0.5; mean average precision was 99.5% and mean average recall was 99.1% for an IoU of 0.5-0.95. We then evaluated our DeepTrac module using video data which is then plot the disease spread directions to understand the pattern of bacterial spot spread in bell-pepper leaf. Our future work may include the application of our AI method to detect plant diseases on the drone-obtained plant images.