DOCSLIB.ORG

Camera Trapping for the Detection of Small Mammals -Trial of Camera Traps to Survey for the New Holland Mouse (Pseudomys Novaehollandiae) in Tasmania

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Camera Trapping for the Detection of Small Mammals -Trial of Camera Traps to Survey for the New Holland Mouse (Pseudomys Novaehollandiae) in Tasmania Natural Values Report for the Natural and Cultural Heritage Division’s Biodiversity Monitoring Program Camera trapping for the detection of small mammals -Trial of camera traps to survey for the New Holland mouse (Pseudomys novaehollandiae) in Tasmania Biodiversity Monitoring Section Micah Visoiu – Biodiversity Monitoring Ecologist Michael Driessen – Senior Zoologist Nature Conservation Report 2018 / 4 Department of Primary Industries, Parks, Water and Environment Natural Values Report for the Natural and Cultural Heritage Division’s Biodiversity Monitoring Program – Camera trapping for the detection of small mammals -Trial of camera traps to survey for the New Holland mouse (Pseudomys novaehollandiae) in Tasmania Project team Micah Visoiu Michael Driessen Acknowledgements Funding for the surveys and report preparation was provided by the Tasmanian government. The authors would like to thank Tim Rudman (NCH) for participating in field work, Rosemary Gales (NCH) and Elise Dewar (NCH) for providing comments on the manuscript and Phoebe Burns of the University of Melbourne who provided useful advice and feedback Copyright 2018 Crown in right of State of Tasmania Apart from fair dealing for the purposes of private study, research, criticism or review, as permitted under the Copyright Act, no part may be reproduced by any means without permission from the Department of Primary Industries, Parks, Water and Environment. Nature Conservation Report 2018/4 ISSN: (electronic): 1838-7403 Published by: Biodiversity Monitoring Section, Natural Values Conservation Branch, Natural and Cultural Heritage Division, Department of Primary Industries, Parks, Water and Environment (DPIPWE), GPO Box 44 Hobart, Tasmania, 7001. Suggested citation: Visoiu, M. and Driessen, M. (2018) Camera trapping for the detection of small mammals -Trial of camera traps to survey for the New Holland mouse (Pseudomys novaehollandiae) in Tasmania, Nature Conservation Report 18/4, Natural and Cultural Heritage Division, Department of Primary Industries, Parks, Water and Environment, Hobart Cover photo: Coastal Heath, Waterhouse Point, Tasmania. i Executive Summary A camera trapping method developed in Victoria for the survey of the threatened New Holland mouse (Pseudomys novaehollandiae) was tested and refined using available equipment. This included the development of a protocol to modify Keepguard 680V camera traps to enhance their suitability for this purpose. New Holland mice have undergone a dramatic decline in recent years in Tasmania, likely as a result of interactions of habitat management practices and climatic conditions. The last unequivocal capture of a New Holland mouse in Tasmania was in 2004, although possible hair samples were collected in 2009 and 2010. In the intervening years substantial trapping effort has been put into re-locating the species both at historically known sites and other sites with suitable habitat. In the process of testing and refining the camera trapping method we conducted a survey of known and potential New Holland mouse sites. Despite over 1200 animal visitations being recorded by the trialled camera traps no images that could be confidently determined as New Holland mice were recorded. Despite this, the camera trapping method modified during the trial is considered to be suitable for wide-scale and/or intensive surveys for this and other small-bodied mammal species in Tasmania. It is far less time and labour intensive than live trapping programs and can be carried out at far greater spatial and temporal scales. This effectively and efficiently increases the likelihood of detection of this highly threatened and cryptic species whilst minimising resources investment. ii Table of Contents Summary ................................................................................................................................................................... ii Table of Contents ................................................................................................................................................. iii 1. Introduction ................................................................................................................................................... 1 2. Methods ........................................................................................................................................................... 2 Site selection ....................................................................................................................................................... 2 Camera trap setup ............................................................................................................................................ 5 Vegetation information..................................................................................................................................... 8 Image identification ........................................................................................................................................... 8 3. Results ............................................................................................................................................................. 8 4. Discussion ..................................................................................................................................................... 11 5. Communication ........................................................................................................................................... 13 6. References .................................................................................................................................................... 14 Appendix one ........................................................................................................................................................ 16 iii 1. Introduction The New Holland mouse (Pseudomys novaehollandiae) has a patchy distribution in coastal eastern Australia from Evans Head in NSW to Angelsea (Victoria), Flinders Island and eastern Tasmania (Menkhorst and Knight 2001). The species was first trapped in Tasmania in 1976 (Hocking 1980) and is likely Tasmania’s rarest terrestrial mammal species. It is listed as endangered under the Tasmanian Threatened Species Protection Act 1995 and vulnerable under the Environmental Protection and Biodiversity Conservation Act 2002. The species was reliably, although sparingly, captured via targeted trapping programs up until the early 2000’s, with several studies undertaken on its biology in the 1980’s and 1990’s. The first of these studies was carried out in the Eddystone Point area of larapuna in 1986-87 (Pye 1991). Subsequently the same population was also studied in 1998-99 along with other sites at Coles Bay, wukalina / Mt William National Park and on Flinders Island (Lazenby et al., 2008). Incidental captures of the species were also made in the Taylors Beach area near The Gardens in 1978 (Green 1979) and in 1982 (Norton 1987a). Norton (1987b) in undertaking a study on the ecology of Swamp Rats (Rattus lutreolus velutinus) in the Gardens area of the Bay of Fires, captured enough New Holland Mice to also complete a small study on their ecology. The last known record of a live capture and thus unequivocal positive record of the species was by Lazenby in Stumpies Bay, wukalina / Mt William National Park in August 2004 (Lazenby unpub. data). A potential hair tube detection was made in 2009 near Binalong Bay and a probable hair tube detection made in 2010 from Waterhouse Point (Lazenby unpub. data). The hair tube detection at Waterhouse Point is the only indication of the species in the area despite a targeted survey with 385 trap nights undertaken in 1993 (Driessen, Hocking and Holdsworth 1995). In the past 10 years there have been targeted (and incidental) trapping surveys at several locations: wukalina / Mt William in 2016 (Wildlife Management Branch unpubl. data), Eddystone Point in 2013 (Biodiversity Monitoring Section unpubl. data), Flinders Island in 2012 (Lazenby unpubl. data), and the Gardens and Binalong Bay areas of larapuna in 2008 and 2009 (Driessen, unpubl. data). These surveys have comprised thousands of trapping nights in locations where New Holland Mice were expected to be present, and would have in the past been expected to be captured. Hair tube surveys have provided some limited success with the aforementioned detections in 2009 and 2010 (Lazenby unpub. Data), and the 2014 DNA in predator scats survey failed to detect the DNA of this species in areas of potential habitat (Modave 2017). The evident reduction in numbers is potentially attributable to management practices, in particular inappropriate fire regimes; the species has been shown to occur in early-mid stages of vegetation succession after fire (Hocking 1980; Lazenby et al., 2008; Seebeck and Menkhorst 2000) and is considered to generally benefit from frequent, small-scale patch burning (Seebeck et al., 1996). However drought has also been shown to significantly impact populations (Lock and Wilson 2017) and prolonged periods of below average rainfall through the 2000’s has likely also contributed to a reduced density of New Holland mice. Given the conservation significance of this species and current uncertainty about its status, the re- location of known populations or locating new populations is integral to the management of critical habitat, and potentially, the long term survival of the species in Tasmania. Due to recent unsuccessful
Load more

Recommended publications
  • A Biological Survey of the Southern Mount Lofty Ranges
    Southern Mount Lofty Ranges Biological Survey APPENDIX I DESCRIPTION OF ENVIRONMENTAL ASSOCIATIONS OCCURRING IN SURVEY REGION BOUNDARY. Part 1. Environmental associations in study area occurring within FLEURIEU IBRA sub-region Environmental Total % of Description Association Area vegetation (ha) remaining 3.2.1 Mt. Rapid 12,763 3.9 Hills and ridges on interbedded shale and arkose, locally overlain by tillite. Relict fans form broad flat surfaces near Cape Jervis where some coastal cliffs occur. Open parkland with sown pasture is used for livestock grazing. The scenery of the coastline is dominated by tall cliffs that vary in form and steepness, the amount of rock outcrop and vegetative cover. 3.2.2 Deep Creek 12,984 30.2 A long dissected ridge of phyllite and greywacke with cliffs, or beaches and dunes along the coastline. The cover is predominantly open parkland over sown pasture with widespread remnants of woodland and forest. Inland views tend to be middle-ground panoramic, featuring grassy ridge crests and valley floors with bracken and reed or remnant forest vegetation. 3.2.3 Fleurieu 30,389 15.6 An undulating to hilly dissected tableland on lateritized sandstone. There is a mixed cover of open parkland, forest plantation and woodland. 3.2.4 Inman 37,130 4.4 A series of low dissected ridges and spurs on tillite and arkose, with dunes and beaches or Valley cliffs along the coast. The cover is open parkland over sown pastures and cereal crops. 3.2.5 Bob Tiers 15,761 21.3 Ridges on schist and gneiss with dissected slopes and remnantsof laterite-capped tableland.
    [Show full text]
  • On the Complexities of Using New Visual Technologies to Communicate About Wildlife Communication
    Verma A, Van der Wal R, Fischer A. Microscope and spectacle: On the complexities of using new visual technologies to communicate about wildlife communication. Ambio 2015, 44(Suppl. 4), 648-660. Copyright: © The Author(s) 2015 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. DOI link to article: http://dx.doi.org/10.1007/s13280-015-0715-z Date deposited: 05/12/2016 This work is licensed under a Creative Commons Attribution 4.0 International License Newcastle University ePrints - eprint.ncl.ac.uk Ambio 2015, 44(Suppl. 4):S648–S660 DOI 10.1007/s13280-015-0715-z Microscope and spectacle: On the complexities of using new visual technologies to communicate about wildlife conservation Audrey Verma, Rene´ van der Wal, Anke Fischer Abstract Wildlife conservation-related organisations well-versed in crafting public outreach and awareness- increasingly employ new visual technologies in their raising activities. These range from unidirectional educa- science communication and public engagement efforts. tional campaigns and advertising and branding projects, to Here, we examine the use of such technologies for wildlife citizen science research with varying degrees of public conservation campaigns. We obtained empirical data from participation, the implementation of interactive media four UK-based organisations through semi-structured strategies, and the expansion of modes of interpretation to interviews and participant observation.
    [Show full text]
  • Jervis Bay Territory Page 1 of 50 21-Jan-11 Species List for NRM Region (Blank), Jervis Bay Territory
    Biodiversity Summary for NRM Regions Species List What is the summary for and where does it come from? This list has been produced by the Department of Sustainability, Environment, Water, Population and Communities (SEWPC) for the Natural Resource Management Spatial Information System. The list was produced using the AustralianAustralian Natural Natural Heritage Heritage Assessment Assessment Tool Tool (ANHAT), which analyses data from a range of plant and animal surveys and collections from across Australia to automatically generate a report for each NRM region. Data sources (Appendix 2) include national and state herbaria, museums, state governments, CSIRO, Birds Australia and a range of surveys conducted by or for DEWHA. For each family of plant and animal covered by ANHAT (Appendix 1), this document gives the number of species in the country and how many of them are found in the region. It also identifies species listed as Vulnerable, Critically Endangered, Endangered or Conservation Dependent under the EPBC Act. A biodiversity summary for this region is also available. For more information please see: www.environment.gov.au/heritage/anhat/index.html Limitations • ANHAT currently contains information on the distribution of over 30,000 Australian taxa. This includes all mammals, birds, reptiles, frogs and fish, 137 families of vascular plants (over 15,000 species) and a range of invertebrate groups. Groups notnot yet yet covered covered in inANHAT ANHAT are notnot included included in in the the list. list. • The data used come from authoritative sources, but they are not perfect. All species names have been confirmed as valid species names, but it is not possible to confirm all species locations.
    [Show full text]
  • Introduction Methods Results
    Papers and Proceedings Royal Society ofTasmania, Volume 1999 103 THE CHARACTERISTICS AND MANAGEMENT PROBLEMS OF THE VEGETATION AND FLORA OF THE HUNTINGFIELD AREA, SOUTHERN TASMANIA by J.B. Kirkpatrick (with two tables, four text-figures and one appendix) KIRKPATRICK, J.B., 1999 (31:x): The characteristics and management problems of the vegetation and flora of the Huntingfield area, southern Tasmania. Pap. Proc. R. Soc. Tasm. 133(1): 103-113. ISSN 0080-4703. School of Geography and Environmental Studies, University ofTasmania, GPO Box 252-78, Hobart, Tasmania, Australia 7001. The Huntingfield area has a varied vegetation, including substantial areas ofEucalyptus amygdalina heathy woodland, heath, buttongrass moorland and E. amygdalina shrubbyforest, with smaller areas ofwetland, grassland and E. ovata shrubbyforest. Six floristic communities are described for the area. Two hundred and one native vascular plant taxa, 26 moss species and ten liverworts are known from the area, which is particularly rich in orchids, two ofwhich are rare in Tasmania. Four other plant species are known to be rare and/or unreserved inTasmania. Sixty-four exotic plantspecies have been observed in the area, most ofwhich do not threaten the native biodiversity. However, a group offire-adapted shrubs are potentially serious invaders. Management problems in the area include the maintenance ofopen areas, weed invasion, pathogen invasion, introduced animals, fire, mechanised recreation, drainage from houses and roads, rubbish dumping and the gathering offirewood, sand and plants. Key Words: flora, forest, heath, Huntingfield, management, Tasmania, vegetation, wetland, woodland. INTRODUCTION species with the most cover in the shrub stratum (dominant species) was noted. If another species had more than half The Huntingfield Estate, approximately 400 ha of forest, the cover ofthe dominant one it was noted as a codominant.
    [Show full text]
  • Plant Press Cross-Border Community Nursery Newsletter Volume 1—Issue 1
    Plant Press Cross-border Community nursery newsletter Volume 1—Issue 1 Jan– Feb 2014 Newsletter debut Welcome to the first edition of the Plant Press Newsletter produced by Nature Glenelg Trusts Cross-Border Community Nursery. The purpose of the Newsletter is to keep growers, seed collectors, revegetators and interested community members up to date with activities and events being run either by the Nursery or in partnership with other organizations. It will also provide current and relevant information about the germination and propagation of native plants discovered by plant scientist and professional and amateur plant growers in our region. Where relevant , knowledge and experience from further afield will be included. Nursery sowed seeds for success in 2013 Volunteers Welcome The Cross-Border Community Nursery invites interested Before arriving in Mount Gambier to take up the position of the Cross- community members join our Friends of the Cross-Border Border Nursery Co-coordinator in late February 2013, I was aware that Nursery Group. there wasn’t much in the way of a nursery to coordinate. Friends Group members will Nonetheless, when I arrived, I headed be involved in all aspects of straight down to Vansittart Park to check propagation, nursery care and out my new work place. After a few laps the development of our com- around the Park I decided that the few munity herbarium. trays of plants placed on top of an old farm We’d love to hear from anyone gate, adjacent to what looked like a derelict Volunteer Samantha Baker who is interested in being in- & Internee, Yvonne Riley volved.
    [Show full text]
  • Vicariance, Climate Change, Anatomy and Phylogeny of Restionaceae
    Botanical Journal of the Linnean Society (2000), 134: 159–177. With 12 figures doi:10.1006/bojl.2000.0368, available online at http://www.idealibrary.com on Under the microscope: plant anatomy and systematics. Edited by P. J. Rudall and P. Gasson Vicariance, climate change, anatomy and phylogeny of Restionaceae H. P. LINDER FLS Bolus Herbarium, University of Cape Town, Rondebosch 7701, South Africa Cutler suggested almost 30 years ago that there was convergent evolution between African and Australian Restionaceae in the distinctive culm anatomical features of Restionaceae. This was based on his interpretation of the homologies of the anatomical features, and these are here tested against a ‘supertree’ phylogeny, based on three separate phylogenies. The first is based on morphology and includes all genera; the other two are based on molecular sequences from the chloroplast genome; one covers the African genera, and the other the Australian genera. This analysis corroborates Cutler’s interpretation of convergent evolution between African and Australian Restionaceae. However, it indicates that for the Australian genera, the evolutionary pathway of the culm anatomy is much more complex than originally thought. In the most likely scenario, the ancestral Restionaceae have protective cells derived from the chlorenchyma. These persist in African Restionaceae, but are soon lost in Australian Restionaceae. Pillar cells and sclerenchyma ribs evolve early in the diversification of Australian Restionaceae, but are secondarily lost numerous times. In some of the reduction cases, the result is a very simple culm anatomy, which Cutler had interpreted as a primitively simple culm type, while in other cases it appears as if the functions of the ribs and pillars may have been taken over by a new structure, protective cells developed from epidermal, rather than chlorenchyma, cells.
    [Show full text]
  • Camera Installation Instructions
    Camera Installation Instructions Before going into the field: The following field equipment and supplies need to be ready and packed before the team leaves for a field assignment: 1) Trail camera unit. Unit should include the actual trail camera with 8 Eneloop batteries and SD card already in place, the trail camera armored case with label attached, and the master python cable lock with matching key on keychain. REMEMBER!!! When packing the camera traps to take in the field do not store them with food or any other substances that might attach odors to the camera trap. Food or other smells might attract animals to the camera trap in the field, thus biasing the estimates of detection probability. 2) GPS capable device. This can either be a portable or handheld GPS unit or a GPS- capable smartphone with either the TopoMaps+ (iPhone) or Backcountry Navigator (Android) application downloaded. 3) Fieldwork folder. The fieldwork folder should include this camera installation sheet; the datasheet required for each check; the station placard to be displayed after a deployment, check, or takedown; a dry erase marker, and a pen. 4) Hiking equipment. Please make sure to have AT LEAST the following whenever going out into the field: a. Daypack and hiking partner (all volunteers MUST be accompanied by AT LEAST one other person b. Map, GPS, or compass c. Extra clothing, food, and water d. First-aid kit e. Headlamp or flashlight f. Matches and fire-starting kit or lighter g. Knife or multi-use camping tool h. Sunglasses and sunscreen i. Water treatment device j.
    [Show full text]
  • Labiatae) in South-Eastern Australia
    319 Contributions to the systematics of Prostanthera (Labiatae) in south-eastern Australia Barry J. Conn Abstract Conn, B.J. (Royal Botanic Gardens, Mrs Macquaries Road, Sydney NSW 2000, Australia) 1997. Contributions to the systematics of Prostanthera (Labiatae) in south-eastern Australia. Telopea 7(4): 319–332. Prostanthera cuneata is regarded as endemic to mainland Australia, with the type material based on cultivated material grown in Formosa garden, Tasmania. The new species Prostanthera galbraithiae B.J.Conn is here formally described. A morphological re-evaluation of Prostanthera melissifolia concludes that this species is endemic to Victoria. Prostanthera melissifolia sensu Conn, from the Northern Tablelands of New South Wales, is P. caerulea R.Br., and Prostanthera sp. aff. melissifolia sensu Beauglehole, from East Gippsland (Victoria), probably is P. incisa R.Br. (s. lat.). A full description, habitat notes, distribution map and illustration of each of the first three species are provided. Introduction This paper presents the conclusions of revisionary studies into the genus Prostanthera section Prostanthera, arising particularly from work towards the preparation of an account of the genus for the forthcoming Flora of Victoria, volume 4. Terminology follows Conn (1984), except inflorescence terminology as modified by Conn (1995). Distribution information for Victoria follows Conn (1993). 1. Distribution of Prostanthera cuneata Benth. Previously, Prostanthera cuneata has been regarded as occurring in New South Wales, Tasmania and Victoria. Although the type specimen of this species (Gunn 725, 7 Dec. 1842) is stated to have been collected from the ‘South Esk’ River, south of Launceston in Tasmania, Buchanan (1988, 1990) states that Gunn actually collected from ‘Formosa’ garden, on the estate of W.E.
    [Show full text]
  • Camera Trapping for Science
    Camera Trapping for Science Richard Heilbrun Tania Homayoun, PhD Texas Parks & Wildlife Texas Parks & Wildlife Session Overview •Introduction to Camera Trapping Basics (1hour) •Hands‐on Practice with Cameras (1.5 hours) •Break (10 min) •Processing Results Presentation & Activity (1 hour 20min) Introduction to Camera Trapping T. Homayoun Overview •Brief history of camera trapping •Various uses of camera trapping •Camera trapping for citizen science: Texas Nature Trackers program •Camera trapping basics •Considerations for the field History of Camera Trapping • 1890s – George Shiras’s trip wire photography of wildlife • 1927 – Frank M Chapman’s photo censusing of Barro Colorado Island, Panama –Starting to document individuals based on markings • 1930s ‐ Tappan George’s North American wildlife photos –Documentation of set‐up, photo processing, logistics George Shiras, 1903, trip wire photo History of Camera Trapping • 1950s – flash bulb replaces magnesium powder –Still using trip wires, often with bait –Recording behaviors & daily activity • 1960s – using beam of light as trip wire –Series of photos, more exposures –6V car batteries to maintain power –Still very heavy set‐ups (~50lbs) • 1970s‐80s – 35mm cameras –Time‐lapse –Much lighter (6‐13 lbs) George Shiras, 1903, trip wire photo History of Camera Trapping • 1990s – infrared trigger system available –Much smaller units –Drastic jump in quality and affordability • Robustness and longevity make more locations and situations accessible to researchers Camera trap records wolves in Chernobyl
    [Show full text]
  • Ecology of Pyrmont Peninsula 1788 - 2008
    Transformations: Ecology of Pyrmont peninsula 1788 - 2008 John Broadbent Transformations: Ecology of Pyrmont peninsula 1788 - 2008 John Broadbent Sydney, 2010. Ecology of Pyrmont peninsula iii Executive summary City Council’s ‘Sustainable Sydney 2030’ initiative ‘is a vision for the sustainable development of the City for the next 20 years and beyond’. It has a largely anthropocentric basis, that is ‘viewing and interpreting everything in terms of human experience and values’(Macquarie Dictionary, 2005). The perspective taken here is that Council’s initiative, vital though it is, should be underpinned by an ecocentric ethic to succeed. This latter was defined by Aldo Leopold in 1949, 60 years ago, as ‘a philosophy that recognizes[sic] that the ecosphere, rather than any individual organism[notably humans] is the source and support of all life and as such advises a holistic and eco-centric approach to government, industry, and individual’(http://dictionary.babylon.com). Some relevant considerations are set out in Part 1: General Introduction. In this report, Pyrmont peninsula - that is the communities of Pyrmont and Ultimo – is considered as a microcosm of the City of Sydney, indeed of urban areas globally. An extensive series of early views of the peninsula are presented to help the reader better visualise this place as it was early in European settlement (Part 2: Early views of Pyrmont peninsula). The physical geography of Pyrmont peninsula has been transformed since European settlement, and Part 3: Physical geography of Pyrmont peninsula describes the geology, soils, topography, shoreline and drainage as they would most likely have appeared to the first Europeans to set foot there.
    [Show full text]
  • Smithsonian Institution Fiscal Year 2021 Budget Justification to Congress
    Smithsonian Institution Fiscal Year 2021 Budget Justification to Congress February 2020 SMITHSONIAN INSTITUTION (SI) Fiscal Year 2021 Budget Request to Congress TABLE OF CONTENTS INTRODUCTION Overview .................................................................................................... 1 FY 2021 Budget Request Summary ........................................................... 5 SALARIES AND EXPENSES Summary of FY 2021 Changes and Unit Detail ........................................ 11 Fixed Costs Salary and Related Costs ................................................................... 14 Utilities, Rent, Communications, and Other ........................................ 16 Summary of Program Changes ................................................................ 19 No-Year Funding and Object-Class Breakout .......................................... 23 Federal Resource Summary by Performance/Program Category ............ 24 MUSEUMS AND RESEARCH CENTERS Enhanced Research Initiatives ........................................................... 26 National Air and Space Museum ........................................................ 28 Smithsonian Astrophysical Observatory ............................................ 36 Major Scientific Instrumentation .......................................................... 41 National Museum of Natural History ................................................... 47 National Zoological Park ..................................................................... 55 Smithsonian Environmental
    [Show full text]
  • Efficient Pipeline for Camera Trap Image Review
    Efficient Pipeline for Camera Trap Image Review Sara Beery Dan Morris Siyu Yang [email protected] [email protected] [email protected] California Institute of Technology Microsoft AI for Earth Microsoft AI for Earth Pasadena, California Redmond, Washington Redmond, Washington Figure 1: Example results from our generic detector, on images from regions and/or species not seen during training. ABSTRACT Previous work has shown good results on automated species Biologists all over the world use camera traps to monitor biodi- classification in camera trap data [8], but further analysis has shown versity and wildlife population density. The computer vision com- that these results do not generalize to new cameras or new geo- munity has been making strides towards automating the species graphical regions [3]. Additionally, these models will fail to recog- classification challenge in camera traps [1, 2, 4–16], but it has proven nize any species they were not trained on. In theory, it is possible difficult to to apply models trained in one region to images collected to re-train an existing model in order to add missing species, but in in different geographic areas. In some cases, accuracy falls off cata- practice, this is quite difficult and requires just as much machine strophically in new region, due to both changes in background and learning expertise as training models from scratch. Consequently, the presence of previously-unseen species. We propose a pipeline very few organizations have successfully deployed machine learn- that takes advantage of a pre-trained general animal detector and ing tools for accelerating camera trap image annotation.
    [Show full text]