Amazon.com

About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Service (business) & Service provider. The organization has 13363 authors who have published 17317 publications receiving 266589 citations.

Learning to Generate Synthetic Data via Compositing

Abstract: We present a task-specific approach to synthetic data generation. Our framework employs a trainable synthesizer network that is optimized to produce meaningful training samples by assessing the strengths and weaknesses of a ‘target’ classifier. The synthesizer and target networks are trained in an adversarial manner wherein each network is updated with a goal to outdo the other. Additionally, we ensure the synthesizer generates realistic data by pairing it with a discriminator trained on real-world images. Further, to make the target classifier invariant to blending artefacts, we introduce these artefacts to background regions of the training images so the target does not over-fit to them. We demonstrate the efficacy of our approach by applying it to different target networks including a classification network on AffNIST [46], and two object detection networks (SSD, Faster-RCNN) on different datasets. On the AffNIST benchmark, our approach is able to surpass the baseline results with just half the training examples. On the VOC person detection benchmark, we show improvements of up to 2.7% as a result of our data augmentation. Similarly on the GMU detection benchmark, we report a performance boost of 3.5% in mAP over the baseline method, outperforming the previous state of the art approaches by as much as 7.5% in individual categories.

Floodplains as an Achilles' heel of Amazonian forest resilience.

Abstract: The massive forests of central Amazonia are often considered relatively resilient against climatic variation, but this view is challenged by the wildfires invoked by recent droughts. The impact of such fires that spread from pervasive sources of ignition may reveal where forests are less likely to persist in a drier future. Here we combine field observations with remotely sensed information for the whole Amazon to show that the annually inundated lowland forests that run through the heart of the system may be trapped relatively easily into a fire-dominated savanna state. This lower forest resilience on floodplains is suggested by patterns of tree cover distribution across the basin, and supported by our field and remote sensing studies showing that floodplain fires have a stronger and longer-lasting impact on forest structure as well as soil fertility. Although floodplains cover only 14% of the Amazon basin, their fires can have substantial cascading effects because forests and peatlands may release large amounts of carbon, and wildfires can spread to adjacent uplands. Floodplains are thus an Achilles' heel of the Amazon system when it comes to the risk of large-scale climate-driven transitions.

Transparency and trust in artificial intelligence systems

Abstract: Assistive technology featuring artificial intelligence (AI) to support human decision-making has become ubiquitous. Assistive AI achieves accuracy comparable to or even surpassing that of human exp...

MOTChallenge: A Benchmark for Single-Camera Multiple Target Tracking

Abstract: Standardized benchmarks have been crucial in pushing the performance of computer vision algorithms, especially since the advent of deep learning. Although leaderboards should not be over-claimed, they often provide the most objective measure of performance and are therefore important guides for research. We present MOTChallenge, a benchmark for single-camera Multiple Object Tracking (MOT) launched in late 2014, to collect existing and new data and create a framework for the standardized evaluation of multiple object tracking methods. The benchmark is focused on multiple people tracking, since pedestrians are by far the most studied object in the tracking community, with applications ranging from robot navigation to self-driving cars. This paper collects the first three releases of the benchmark: (i) MOT15, along with numerous state-of-the-art results that were submitted in the last years, (ii) MOT16, which contains new challenging videos, and (iii) MOT17, that extends MOT16 sequences with more precise labels and evaluates tracking performance on three different object detectors. The second and third release not only offers a significant increase in the number of labeled boxes, but also provide labels for multiple object classes beside pedestrians, as well as the level of visibility for every single object of interest. We finally provide a categorization of state-of-the-art trackers and a broad error analysis. This will help newcomers understand the related work and research trends in the MOT community, and hopefully shed some light into potential future research directions.

The Timing of Staffing Decisions in Hospital Operating Rooms: Incorporating Workload Heterogeneity into the Newsvendor Problem

Abstract: We study the problem of setting nurse staffing levels in hospital operating rooms when there is uncertainty about daily workload. The workload is the number of operating room hours used by a medical specialty on a given day to perform surgical procedures. Variable costs consist of wages at a regular (scheduled) rate and at an overtime rate when the realized workload exceeds the scheduled time. Using a newsvendor framework, we consider the problem of determining optimal staffing levels with different information sets available at the time of decision: no information, information on number of cases, and information on number and types of cases. We develop empirical models for the daily workload distribution in which the mean and variance change with the information available. We use these models to derive optimal staffing rules based on historical data from a U.S. teaching hospital and prospectively test the performance of these rules. Our numerical results suggest that hospitals could potentially reduce their staffing costs by up to 39%--49% by deferring staffing decisions until procedure type information is available. The results demonstrate how data availability can affect a newsvendor's performance. The systematic approach of empirical modeling presented in the paper can be applied to other newsvendor problems with heterogeneous sources of demand.