The Grid 2: Blueprint for a New Computing Infrastructure

Having accurate, detailed, and up-to-date information about the location and behavior of animals in the wild would improve our ability to study and conserve ecosystems. We investigate the ability to automatically, accurately, and inexpensively collect such data, which could help catalyze the transformation of many fields of ecology, wildlife biology, zoology, conservation biology, and animal behavior into “big data” sciences. Motion-sensor “camera traps” enable collecting wildlife pictures inexpensively, unobtrusively, and frequently. However, extracting information from these pictures remains an expensive, time-consuming, manual task. We demonstrate that such information can be automatically extracted by deep learning, a cutting-edge type of artificial intelligence. We train deep convolutional neural networks to identify, count, and describe the behaviors of 48 species in the 3.2 million-image Snapshot Serengeti dataset. Our deep neural networks automatically identify animals with >93.8% accuracy, and we expect that number to improve rapidly in years to come. More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving >8.4 y (i.e., >17,000 h at 40 h/wk) of human labeling effort on this 3.2 million-image dataset. Those efficiency gains highlight the importance of using deep neural networks to automate data extraction from camera-trap images, reducing a roadblock for this widely used technology. Our results suggest that deep learning could enable the inexpensive, unobtrusive, high-volume, and even real-time collection of a wealth of information about vast numbers of animals in the wild.

https://www.pnas.org/content/pnas/115/25/E5716.full.pdf

Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning.

Having accurate, detailed, and up-to-date information about the location and behavior of animals in the wild would revolutionize our ability to study and conserve ecosystems. We investigate the ability to automatically, accurately, and inexpensively collect such data, which could transform many fields of biology, ecology, and zoology into "big data" sciences. Motion sensor "camera traps" enable collecting wildlife pictures inexpensively, unobtrusively, and frequently. However, extracting information from these pictures remains an expensive, time-consuming, manual task. We demonstrate that such information can be automatically extracted by deep learning, a cutting-edge type of artificial intelligence. We train deep convolutional neural networks to identify, count, and describe the behaviors of 48 species in the 3.2-million-image Snapshot Serengeti dataset. Our deep neural networks automatically identify animals with over 93.8% accuracy, and we expect that number to improve rapidly in years to come. More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving more than 8.4 years (at 40 hours per week) of human labeling effort (i.e. over 17,000 hours) on this 3.2-million-image dataset. Those efficiency gains immediately highlight the importance of using deep neural networks to automate data extraction from camera-trap images. Our results suggest that this technology could enable the inexpensive, unobtrusive, high-volume, and even real-time collection of a wealth of information about vast numbers of animals in the wild.

Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning

Global change is impacting forests worldwide, threatening biodiversity and ecosystem services including climate regulation. Understanding how forests respond is critical to forest conservation and climate protection. This review describes an international network of 59 long-term forest dynamics research sites (CTFS-ForestGEO) useful for characterizing forest responses to global change. Within very large plots (median size 25ha), all stems 1cm diameter are identified to species, mapped, and regularly recensused according to standardized protocols. CTFS-ForestGEO spans 25 degrees S-61 degrees N latitude, is generally representative of the range of bioclimatic, edaphic, and topographic conditions experienced by forests worldwide, and is the only forest monitoring network that applies a standardized protocol to each of the world's major forest biomes. Supplementary standardized measurements at subsets of the sites provide additional information on plants, animals, and ecosystem and environmental variables. CTFS-ForestGEO sites are experiencing multifaceted anthropogenic global change pressures including warming (average 0.61 degrees C), changes in precipitation (up to +/- 30% change), atmospheric deposition of nitrogen and sulfur compounds (up to 3.8g Nm(-2)yr(-1) and 3.1g Sm(-2)yr(-1)), and forest fragmentation in the surrounding landscape (up to 88% reduced tree cover within 5km). The broad suite of measurements made at CTFS-ForestGEO sites makes it possible to investigate the complex ways in which global change is impacting forest dynamics. Ongoing research across the CTFS-ForestGEO network is yielding insights into how and why the forests are changing, and continued monitoring will provide vital contributions to understanding worldwide forest diversity and dynamics in an era of global change.

/pdf/ctfs-forestgeo-a-worldwide-network-monitoring-forests-in-an-2z2d2x7id0.pdf

CTFS-ForestGEO: A worldwide network monitoring forests in an era of global change

Camera traps can be used to address large-scale questions in community ecology by providing systematic data on an array of wide-ranging species. We deployed 225 camera traps across 1,125 km(2) in Serengeti National Park, Tanzania, to evaluate spatial and temporal inter-species dynamics. The cameras have operated continuously since 2010 and had accumulated 99,241 camera-trap days and produced 1.2 million sets of pictures by 2013. Members of the general public classified the images via the citizen-science website www.snapshotserengeti.org. Multiple users viewed each image and recorded the species, number of individuals, associated behaviours, and presence of young. Over 28,000 registered users contributed 10.8 million classifications. We applied a simple algorithm to aggregate these individual classifications into a final 'consensus' dataset, yielding a final classification for each image and a measure of agreement among individual answers. The consensus classifications and raw imagery provide an unparalleled opportunity to investigate multi-species dynamics in an intact ecosystem and a valuable resource for machine-learning and computer-vision research.

/pdf/snapshot-serengeti-high-frequency-annotated-camera-trap-4lwzu17hr5.pdf

Snapshot Serengeti, high-frequency annotated camera trap images of 40 mammalian species in an African savanna

Data acquisition and management software for camera trap data: A case study from the TEAM Network

We have developed the GEONGrid system for coordinating and managing naturally distributed computing, data, and cluster resources on the cyberinfrastructure. Recently, since the use of Grid technology is still very complex for researchers and scientists, the area of Grid Portals has made excellent progress. The Grid portal system is an emerging open Grid computing environment that promises to provide users with uniform seamless access to remote computing and data resources by providing an easy to use interface to cover over the complexity of more sophisticated Grid technologies. In this paper, we present our initial efforts in the design and implementation of service components in the GEONGrid portal. These service components may be implemented as Web services that follow the conventions of service-oriented architecture design. In this approach, service components are self-contained, have a well-defined programming interface defined in WSDL, and communicate using SOAP messaging. In building a GEONGrid portal, we also use a component-based user interface design. Portlets provide the desired component model for user interfaces in the same way as Web services. Using this approach, which allows Grid portals to be built out of reusable components, has the obvious advantages of reusability and modularity. Copyright © 2007 John Wiley & Sons, Ltd.

GEONGrid portal: design and implementations

The GEOsciences Network (GEON, ) is a large-scale collaborative cyberinfrastructure project involving information technology and geoscience researchers from multiple institutions. The GEON infrastructure provides portal, middleware, and data resources to facilitate scientific discovery for domain scientists using applications, tools, and services. It consists of both a service-oriented Web-Grid framework and application toolkits, using the Web service and portlet programming model to represent applications. Based on those grid environments, we have developed the SYNSEIS (SYNthetic SEISmogram) tool within the GEON infrastructure to support personalized experiments in seismology. In this paper, we present an overview of SYNSEIS from a user point of view, and demonstrate how one can use a simple management scheme to perform a parameter sweep and distribute the work in computational resources, using a scientific application that was not specifically designed to perform parameter sweeps. The performance advantages to be gained by using this scheme with scientific codes for dealing with a large number of jobs on computational grids are very substantial. In particular, we identify the earthquake simulations in the SYNSEIS tool as an example application that can benefit from running jobs on computational resources and subsequently promote the sharing of computational resources among partner sites involved in the GEON project. Finally, we also discuss the parallel scaling behavior of our primary earthquake simulation application. Copyright © 2010 John Wiley & Sons, Ltd.

Management of a parameter sweep for scientific applications on cluster environments

Leveraging service-oriented architectures and taking advantage of the high-performance compute resources provided by XSEDE, we have developed standards-based web services to address the challenges associated with processing large volumes of high resolution topography data. These web services make results from community software packages and other cyberinfrastructure-based applications available to the wider earth sciences community via the OpenTopography Facility and the CyberGIS Gateway.

Leveraging XSEDE HPC resources to address computational challenges with high-resolution topography data

The Geosciences Network (GEON) project has been developing cyberinfrastructure for data sharing in the Earth Science community based on a service-oriented architecture. The layered architecture consists of Core, Middleware, and Applications services. Core services provide system-level functions (e.g. user authentication), Middleware services provide generic capabilities (e.g. catalog search), and Application services provide functions that users directly interact with, including applications that are specific to Earth Sciences. The GEON ‘service stack’ includes a standardized set of these services and the corresponding software modules. The GEON Portal provides Web-based access to these services via a set of portlets. This service-oriented approach has enabled GEON to expand to new partner sites and leverage GEON services for other projects. To facilitate interoperation in a distributed geoinformatics environment, GEON is focusing on standards for distributed search across federated catalogs.

Choonhan Youn

Papers

Data acquisition and management software for camera trap data: A case study from the TEAM Network

GEONGrid portal: design and implementations

Management of a parameter sweep for scientific applications on cluster environments

Leveraging XSEDE HPC resources to address computational challenges with high-resolution topography data

The GEON service-oriented architecture for Earth Science applications