Tim,

Hi.  the easy and short answer to that question is Wildlife Acoustics SM4bat or EMtouch, with Kaleidoscope Pro software. The longer answer is.. it depends, and I'm happy to chat more if you like.  You'll need to decide between heterodyne, frequency-division, zero-crossing, time expansion or full-spectrum recording hardware systems.  Automated or handheld? And do you want to build your own species recognizers,  identify calls manually, or rely on existing proprietary (and paid for) identification software? 

Thanks, Carlos

Hi Andy,

My name is Elizabeth (Liz) Bondi, I'm a PhD Candidate in Computer Science at Harvard John A. Paulson School of Engineering and Applied Sciences, advised by Prof. Milind Tambe. In brief, my research has primarily focused on using drones and AI for wildlife conservation.

As an intern at Microsoft Research with Dr. Ashish Kapoor, Dr. Debadeepta Dey, and Dr. Lucas Joppa, we built a simulated African savanna environment in Microsoft AirSim, and flew a simulated drone around it. Our goal was to create an automatically-labeled image dataset for training machine learning models, but I have also used it for demonstrations in the past to allow students to find specific animals while flying the drone around, for example. The AirSim environment will likely require GPUs, but perhaps access to the cloud (e.g., Microsoft Azure, Amazon AWS) could be purchased for students.

One other note, I believe Sara Beery also used AirSim for camera traps.

Please feel free to check out Microsoft AirSim and download the African savanna environment (download "Africa.zip"). Our papers (see "BIRDSAI" 2020 and "AirSim-W" 2018 papers in particular) and the dataset are also available for some more information. Please feel free to email me at [email protected] if you would like to discuss anything further.

Best,

Liz

Hi Tom, 

okay good news, Aaron's been able to track down the designs for you. He's going to drop them in the slack chat - he'll be the person you can ask questions if needed as well. 
Steph 

Hi Steph - as a career technologist now looking to see how I can engage my skills and experience in the world of conservation technology, this is pretty interesting review of the state of play right now.

All organisations (commercial, government, not-for-profits, military and other services ) struggle with the question of how and when to deploy technology as part of their overall systems of operation and how then to ensure that they reap the expected benefits. It's not simple, there is rarely if ever a perfect solution of any scope and longevity but of course many organisations do garner very significant beneifts from the efficient use of mainstream technologies and the innovative use of unusual or emerging technologies.

A few quick remarks spring to mind from your notes:

1) The Gartner hype cycle - it's a pretty useful visualisation (much of the latter part of my career was spent trying to guide technologies targeted at large enterprises from the "early adopter" to the "mainstream" stages) and most useful for helping a tech user to decide what kind of user they should be in their current context and where the technologies that are coming under consideration fall on that cycle of maturity. A mismatch ? Then take care ...

2) Conservation tech is not a washing machine - well that depends. I imagine there are a number of different techs and projects and the "washing machine" status will vary. If you know that the tech you NEED to meet your project aims is unproven, don't expect a washing machine level of ease of use !  If you think you're deploying a mainstream tech, then DO expect a "washing machine". As an example, if you buy and deploy a simple camera trap, you expect a good manual (albeit developed for the manufacturer's target market, which may be hunters in the USA)  and you expect the device to operate in accordance with the manual. If you're deploying a set of networked sensors that will use AI in real time to detect and interact with animals in the environment - don't expect a smooth ride. In fact in this latter case, proving the technology may be one of the objectives of the study, whereas in the former, the data being collected was the required resource for the study objectives. 

3) System engineering and enterprise architecture are the two disciplines that organisation use to try to get their whole tech strategy and execution to work to meet their goals. Done well, it means that the business goals are identified and understood and the allocation of tasks between people and different techs is optimally made for performance and cost effectiveness. It's how to ensure a well balanced overall system and process approach. It's an area that been studied and there are methods for implementing this discipline. In fact, if I then add there are many, many methods for implementation, you may realise that it's a problem that's easily stated and less easily solved, but still is generally better that having a splash on the latest shiny tech and hoping for the best.

4) There will be trends, experiences and best practices that can be shared, but naturally these emerge as technologies mature - nobody knows best practice the first time  and innovators will also necessarily face unique challenges. However defining some ontologies or frameworks to help record and structure experiences may very well be worthwhile

Andy 

Oh yes I forgot to mention, I will be at the University of Nottingham campus on Saturday too.

Unfortunately, we're not set up to enable download for video recordings at the moment, but we do plan to upload shorter clips of each presentation to YouTube in the near future. We could also potentially make an audio-only version available for download if that would be of interest?