Animal recognition and identification with deep convolutional neural networks for automated wildlife monitoring

Authors: Hung Nguyen, Sarah J Maclagan, Tu Dinh Nguyen, Thin Nguyen, Paul Flemons, Kylie Andrews, Euan G Ritchie, and Dinh Phung

Published in: 2017 IEEE International Conference on Data Science and Advanced Analytics

Abstract

Efficient and reliable monitoring of wild animals in their natural habitats is essential to inform conservation and management decisions. Automatic covert cameras or “camera traps” are being an increasingly popular tool for wildlife monitoring due to their effectiveness and reliability in collecting data of wildlife unobtrusively, continuously and in large volume. However, processing such a large volume of images and videos captured from camera traps manually is extremely expensive, time-consuming and also monotonous. This presents a major obstacle to scientists and ecologists to monitor wildlife in an open environment.

Leveraging on recent advances in deep learning techniques in computer vision, we propose in this paper a framework to build automated animal recognition in the wild, aiming at an automated wildlife monitoring system. In particular, we use a single-labeled dataset from Wildlife Spotter project, done by citizen scientists, and the state-of-the-art deep convo- lutional neural network architectures, to train a computational system capable of filtering animal images and identifying species automatically.

Our experimental results achieved an accuracy at 96.6% for the task of detecting images containing animal, and 90.4% for identifying the three most common species among the set of images of wild animals taken in South-central Victoria, Australia, demonstrating the feasibility of building fully automated wildlife observation. This, in turn, can therefore speed up research findings, construct more efficient citizen science- based monitoring systems and subsequent management decisions, having the potential to make significant impacts to the world of ecology and trap camera images analysis.

Nguyen H, Maclagan SJ, Nguyen TD, Nguyen T, Flemons P, Andrews K, Ritchie EG, Phung D (2017) Animal recognition and identification with deep convolutional neural networks for automated wildlife monitoring, 2017 IEEE International Conference on Data Science and Advanced Analytics PDF DOI 

From drone swarms to tree batteries, new tech is revolutionising ecology and conservation

Eyes in the sky: drone footage is becoming a vital tool for monitoring ecosystems. Image credit: Deakin Marine Mapping Group

By Euan Ritchie and Blake Allen, Deakin University

This article was originally published on The Conversation. Read the original article, including reader comments.

Understanding Earth’s species and ecosystems is a monumentally challenging scientific pursuit. But with the planet in the grip of its sixth mass extinction event, it has never been a more pressing priority.

To unlock nature’s secrets, ecologists turn to a variety of scientific instruments and tools. Sometimes we even repurpose household items, with eyebrow-raising results – whether it’s using a tea strainer to house ants, or tackling botfly larvae with a well-aimed dab of nail polish.

But there are many more high-tech options becoming available for studying the natural world. In fact, ecology is on the cusp of a revolution, with new and emerging technologies opening up new possibilities for insights into nature and applications for conserving biodiversity.

Our study, published in the journal Ecosphere, tracks the progress of this technological development. Here we highlight a few examples of these exciting advances.

Tiny tracking sensors

Electronically recording the movement of animals was first made possible by VHF radio telemetry in the 1960s. Since then even more species, especially long-distance migratory animals such as caribou, shearwaters and sea turtles, have been tracked with the help of GPS and other satellite data.

But our understanding of what affects animals’ movement and other behaviours, such as hunting, is being advanced further still by the use of “bio-logging” – equipping the animals themselves with miniature sensors.

Many types of miniature sensors have now been developed, including accelerometers, gyroscopes, magnetometers, micro cameras, and barometers. Together, these devices make it possible to track animals’ movements with unprecedented precision. We can also now measure the “physiological cost” of behaviours – that is, whether an animal is working particularly hard to reach a destination, or within a particular location, to capture and consume its prey.

Taken further, placing animal movement paths within spatially accurate 3D-rendered (computer-generated) environments will allow ecologists to examine how individuals respond to each other and their surroundings.

These devices could also help us determine whether animals are changing their behaviour in response to threats such as invasive species or habitat modification. In turn, this could tell us what conservation measures might work best.

Autonomous vehicles

Remotely piloted vehicles, including drones, are now a common feature of our skies, land, and water. Beyond their more typical recreational uses, ecologists are deploying autonomous vehicles to measure environments, observe species, and assess changes through time, all with a degree of detail that was never previously possible.

Coupling autonomous vehicles with sensors (such as thermal imaging) now makes it easier to observe rare, hidden or nocturnal species. It also potentially allows us to catch poachers red-handed, which could help to protect animals like rhinoceros, elephants and pangolins.

3D printing

Despite 3D printing having been pioneered in the 1980s, we are only now beginning to realise the potential uses for ecological research. For instance, it can be used to make cheap, lightweight tracking devices that can be fitted onto animals. Or it can be used to create complex and accurate models of plants, animals or other organisms, for use in behavioural studies.

Bio-batteries

Keeping electronic equipment running in the field can be a challenge. Conventional batteries have limited life spans, and can contain toxic chemicals. Solar power can help with some of these problems, but not in dimly lit areas, such as deep in the heart of rainforests.

“Bio-batteries” may help to overcome this challenge. They convert naturally occurring sources of chemical energy, such as starch, into electricity using enzymes. “Plugging-in” to trees may allow sensors and other field equipment to be powered cheaply for a long time in places without sun or access to mains electricity.

Combining technologies

All of the technologies described above sit on a continuum from previous (now largely mainstream) technological solutions, to new and innovative ones now being trialled.

Emerging technologies are exciting by themselves, but when combined with one another they can revolutionise ecological research. Here is a modified exerpt from our paper:

These advancements will not only generate more accurate research data, but should also minimise the disturbance to species and ecosystems in the process.

Not only will this minimise the stress to animals and the inadvertent spread of diseases, but it should also provide a more “natural” picture of how plants, animals and other organisms interact.

Realising the techno-ecological revolution will require better collaboration across disciplines and industries. Ecologists should ideally also be exposed to relevant technology-based training (such as engineering or IT) and industry placements early in their careers.

The ConversationSeveral initiatives, such as Wildlabs, the Conservation Technology Working Group and TechnEcology, are already addressing these needs. But we are only just at the start of what’s ultimately possible.
The Conversation

Futurecasting ecological research: the rise of technoecology

Authors: Blake M Allan, Dale G Nimmo, Daniel Ierodiaconou, Jeremy VanDerWal, Lian Pin Koh, and Euan G Ritchie

Published in: Ecosphere, volume 9, issue 5 (May 2018)

Abstract

Increasingly complex research questions and global challenges (e.g., climate change and biodiversity loss) are driving rapid development, refinement, and uses of technology in ecology. This trend is spawning a distinct sub‐discipline, here termed “technoecology.”

We highlight recent ground‐breaking and transformative technological advances for studying species and environments: bio‐batteries, low‐power and long‐range telemetry, the Internet of things, swarm theory, 3D printing, mapping molecular movement, and low‐power computers. These technologies have the potential to revolutionize ecology by providing “next‐generation” ecological data, particularly when integrated with each other, and in doing so could be applied to address a diverse range of requirements (e.g., pest and wildlife management, informing environmental policy and decision making).

Critical to technoecology’s rate of advancement and uptake by ecologists and environmental managers will be fostering increased interdisciplinary collaboration. Ideally, such partnerships will span the conception, implementation, and enhancement phases of ideas, bridging the university, public, and private sectors.

Allan BM, Nimmo DG, Ierodiaconou D, VanDerWal J, Koh LP, Ritchie EG (2018) Futurecasting ecological research: the rise of technoecology, Ecosphere PDF DOI

Planet Earth II Live unites art and science in a celebration of nature

Flamingoes dance on a lake in South America in Planet Earth II Live in Concert. Image credit: Travis Hayto

By Euan Ritchie, Deakin University

This article was originally published on The Conversation. Read the original article, including reader comments.

Following an epic and determined journey by an amorous male pygymy three-toed sloth, witnessing golden eagles duel over a fox carcass against a backdrop of majestic mountains, or simply being in awe of one of evolution’s most sublime creations, the sword-billed hummingbird, it’s fair to say I was more than a little excited to experience the Melbourne production of BBC’s Planet Earth II Live in Concert.

The show is a fusion of BBC’s extraordinary wildlife and landscape footage, from the Planet Earth II documentary series. It’s presented on a suitably massive screen, accompanied by a live orchestra playing a score written specially for this visual and aural celebration of nature.

The start of the show was sensational, a montage of wildlife imagery from around the world, with the emotional roller-coaster expertly enhanced by the beauty and fusion with the music that filled the room. It was akin to walking through a gallery, where the individual, finer details of paintings were unimportant, but the overall majesty of what was on show swept you away. I wish the remainder of the show had continued in this vein.

The actual images of wildlife were as we’ve come to expect from the BBC’s legendary wildlife team, second to none. A close-up of a lioness’s paw rippling as it moved across the sands of the Namib Desert while stalking a giraffe, a starling murmuration above Rome, langurs leaping through Jodhpur and, of course, the now famous great escape by a hatchling marine iguana from scores of hungry racer snakes were all captivating.

The orchestra was outstanding. Such was the power of the wildlife footage that at times it was possible to forget their presence, but at key moments onscreen they were most certainly heard and felt, providing a truly emotional and visceral experience.

There were many things to like, but a number of important elements didn’t work, which really detracted from what the show could have been. The first and perhaps most important problem was that each sequence was introduced, in far too much detail, by Eric Bana, who had the unenviable task of trying to replace some of David Attenborough’s narration of the original documentary series.

Bana’s commentary was aimed at making people more informed about the wildlife being featured, but it had the unfortunate effect of giving away exactly what was about to happen next, rather than allowing the audience to have the joy of discovering this for themselves, a very odd production decision. It made the performance feel quite stilted and disjointed, which wasn’t helped by an unannounced 20-minute intermission mid-show.

Bana is a fine Australian actor, but his casual approach and lighthearted jokes were unnecessary, and often not particularly funny. Most importantly, it took attention away from the main event, which should have been what was happening on screen. It reminded me of watching a great film on commercial TV and being engrossed, only to be wrenched out of this blissful state by the dreaded commercial break. This happened multiple times throughout the evening. Judicious and sparing use of surtitles could have remedied this situation.

Another deficiency was the editing and sequence of footage. We were told about the plight of a primate relative of humans, the Indri, from Madagascar’s rapidly disappearing forests. So why then did this sequence start with flying draco lizards from Southeast Asia and toucans from South America? Perhaps many were not bothered by this and instead it just reflects my ecological background.

The Indri was one of several species to appear a number of times in the show, sometimes without narration or explanation. This seemed puzzling given how much amazing material the BBC has at its disposal. Australian wildlife was also conspicuous by its absence.

Finally, at times the show’s lighting was over-the-top, washing out and obscuring imagery on the screen.

There was an attempt towards the end of the show to send a message to the audience about the plight of life of Earth, including how our cities could better accommodate plants and animals to live alongside us. There were painful times during the show when I felt like I was watching soon-to-be ghosts, given the current and dire mass extinction event we have created and are witnessing. I couldn’t help notice the irony of the many audience members watching wildlife while drinking bottled water they’d bought during the show’s intermission, a symptom of our dependence on plastics and consumption.

There is no question in my mind that a greater union between the arts and sciences, such as this event, has enormous potential for positive change.

So my greatest hope is that everyone who attended the show felt moved the next day to assess their own choices and how they affect the other species with which we share this planet, and that together we collectively demand our governments to ensure a sustainable future that preserves Earth’s remarkable wonders.
The Conversation

The Conversation: Rockin’ the suburbs – bandicoots live among us in Melbourne

Young southern brown bandicoots (Isoodon obesulus obesulus), an endangered marsupial species living in outer Melbourne. Image credit: Sarah Maclagan.

By Euan Ritchie and Sarah Maclagan, Deakin University.

This article was originally published on The Conversation. Read the original article, including reader comments.

There are more and more of us on Earth, and increasingly we’re choosing to live in cities. This is a problem for wildlife: urbanisation is one of the greatest threats to biodiversity. Sprawling construction to accommodate people completely removes or modifies the homes of many other species.

But although we need to do all we can to mitigate these effects and their causes, cities are far from biodiversity wastelands. In fact, cities are home to large numbers of species, many of which are threatened.

One such example is the iconic but endangered southern brown bandicoot. New research has found these shy animals thriving in peri-urban areas (the interface between cities and more rural areas) on the outskirts of Melbourne.

Novel ecosystems

Through our effects on the environment, humans are increasingly creating “novel ecosystems”: areas composed of new combinations of species and/or new and modified environmental conditions. In these areas there are winners and losers. How we perceive and manage such ecosystems could have a big influence on conserving species more broadly, and helping to address Earth’s extinction crisis.

We studied these issues by examining the ecology of the endangered southern brown bandicoot in southeast Melbourne. Specifically, we compared bandicoot abundance and body condition between novel sites, made up of peri-urban areas, and more natural remnant sites at the Royal Botanic Gardens at Cranbourne and Quail Island Nature Conservation Reserve, which more closely resemble original bandicoot habitat.

Our goal was to see whether novel sites could support resident bandicoots. Crucially, could peri-urban areas allow bandicoots to breed, recruit (attract new migrating individuals) and survive well enough for populations to persist?

Don’t judge habitats by their appearance

Contrary to what the Human Threat Hypothesis (and logic) might predict, we found more bandicoots at novel sites (66 individuals recorded over 1722 trap-nights) than remnant sites (26 individuals recorded over 1384 trap-nights). And bandicoots were most abundant at the novel site with the most urbanised surroundings. The condition of females was similar between novel and remnant sites.

Most bandicoots at novel sites were resident (meaning they were observed at the same location multiple times thoughout the study). At these same sites we recorded successful breeding, recruitment of young adults, and survival of mature adults.

Our results challenge conventional conservation thinking. Where bandicoots did best is also where known predators such as foxes and feral cats are present and abundant, as opposed to the nature reserves from which they are largely absent. Remnant areas also have more intact native vegetation, whereas bandicoots in urban areas nested in roadsides full of invasive blackberry, a weed often targeted for removal.

Bandicoots like thick vegetation, but they appear not to care which plant species they use – as long as cover is sufficient. Blackberry bushes may protect bandicoots from predators, and also be a source of food through the insects they attract.

On the topic of food and just how adaptable bandicoots are, locals within our study region reported them dining on pet and domestic animal food from backyards.

Conservation opportunities in cities

A growing body of evidence suggests that even heavily modified environments can support viable populations of native plants and animals, and we should endeavour to manage these areas more sympathetically for the benefit of more species. This does not mean that all species will thrive in cities and heavily modified environments – there will always be a need for conservation reserves – but it throws the wilderness-versus-city dichotomy into question.

We need greater awareness of the nature we already share our cities with. Deliberate encouragement such as greener building and urban design that encourages wildlife to return and flourish would substantially benefit humans and other species alike. We are a part of and dependent upon nature, and as such should celebrate and seek to re-establish these vital connections.
The Conversation

Don’t judge habitat on its novelty: Assessing the value of novel habitats for an endangered mammal in a peri-urban landscape

Authors: Sarah J Maclagan, Terry Coates, and Euan G Ritchie

Published in: Biological Conservation, volume 223 (July 2018)

Abstract

Novel ecosystems are increasingly common worldwide, particularly in areas heavily impacted by humans such as urban and peri-urban landscapes. Consequently, interest in their potential contribution to biodiversity conservation is growing, including their ability to sustain populations of threatened species. However, few studies have explored whether novel habitats can support viable populations over time and how they compare to less modified, remnant habitats.

We investigated the capacity for novel habitats to support an endangered mammal, the southern brown bandicoot (Isoodon obesulus obesulus: Peramelidae), in a highly-modified landscape near Australia’s second largest city, Melbourne. We compared bandicoot abundance and body condition between five novel and two remnant sites, and examined whether novel sites support residency and key demographic processes necessary for bandicoot population persistence. We found that bandicoot abundance was higher at novel than remnant sites, with the highest abundance at the novel site with the most urbanised surroundings. Female body condition was similar between novel and remnant sites. The majority of bandicoots at novel sites were resident, and breeding activity, recruitment of first-year adults, and survival of mature adults were observed at all novel sites.

Our results demonstrate the potential significance of novel habitats for conserving threatened species within heavily-modified landscapes, and encourage us not to judge the quality of habitats on their novelty alone. Broadening our appreciation of the potential value of novel ecosystems could increase off-reserve species conservation opportunities, a key priority within the context of the Anthropocene and unprecedented global change and biodiversity loss.

Maclagan SJ, Coates T, Ritchie EG (2018) Don’t judge habitat on its novelty: Assessing the value of novel habitats for an endangered mammal in a peri-urban landscape, Biological Conservation PDF DOI 

Testing the assumptions of the pyrodiversity begets biodiversity hypothesis for termites in semi-arid Australia

Authors: Hayley Davis, Euan G Ritchie, Sarah Avitabile, Tim Doherty, and Dale G Nimmo

Published in: The Royal Society Open Science (April 2018)

Abstract

Fire shapes the composition and functioning of ecosystems globally. In many regions, fire is actively managed to create diverse patch mosaics of fire-ages under the assumption that a diversity of post-fire-age classes will provide a greater variety of habitats, thereby enabling species with differing habitat requirements to coexist, and enhancing species diversity (the pyrodiversity begets biodiversity hypothesis). However, studies provide mixed support for this hypothesis.

Here, using termite communities in a semi-arid region of southeast Australia, we test four key assumptions of the pyrodiversity begets biodiversity hypothesis:

  1. that fire shapes vegetation structure over sufficient time frames to influence species’ occurrence,
  2. that animal species are linked to resources that are themselves shaped by fire and that peak at different times since fire,
  3. that species’ probability of occurrence or abundance peaks at varying times since fire, and
  4. that providing a diversity of fire-ages increases species diversity at the landscape scale.

Termite species and habitat elements were sampled in 100 sites across a range of fire-ages, nested within 20 landscapes chosen to represent a gradient of low to high pyrodiversity. We used regression modelling to explore relationships between termites, habitat and fire.

Fire affected two habitat elements (coarse woody debris and the cover of woody vegetation) that were associated with the probability of occurrence of three termite species and overall species richness, thus supporting the first two assumptions of the pyrodiversity hypothesis. However, this did not result in those species or species richness being affected by fire history per se. Consequently, landscapes with a low diversity of fire histories had similar numbers of termite species as landscapes with high pyrodiversity.

Our work suggests that encouraging a diversity of fire-ages for enhancing termite species richness in this study region is not necessary.

Davis H, Ritchie EG, Avitabile S, Doherty T, Nimmo DG (2018) Testing the assumptions of the pyrodiversity begets biodiversity hypothesis for termites in semi-arid Australia, Royal Society Open Science PDF DOI