WILDLABS logoWildlabs.net : The conservation technology network

WILDLABS logoWildlabs.net : The conservation technology network

Group

AI for Conservation / Feed

Artificial intelligence is increasingly being used in the field to analyse information collected by wildlife conservationists, from camera trap and satellite images to audio recordings. AI can learn how to identify which photos out of thousands contain rare species; or pinpoint an animal call out of hours of field recordings - hugely reducing the manual labour required to collect vital conservation data.

discussion

Wildlife data from autonomous vehicles

Riley Knoedler

16 October 2024 7:57pm

It occurred to me and my colleagues recently that the self-driving vehicle industry must be working on their own animal classification models, because it matters if you are about...

1 0
Brett Muir

17 October 2024 5:59am

Comma.ai would be your best bet. Not sure if it's officially open source but was started by a noteable hacker with an open source ethos. Could try reaching out to them.

See full post
discussion

Deepfaune v1.2!

Simon Chamaillé-Jammes

4 October 2024 4:12pm

 We have just released Deepfaune v1.2!! Deepfaune is a freely available software allowing to automatically classify species in camera-traps pictures or videos collected in...

2 4
Jennifer Zhuge

9 October 2024 12:01am

Edit: SOLVED Thanks!

Thank you so much for this awesome work! I was trying to load the v2 model the same way as in classifTools.py:
model = timm.create_model(backbone="vit_large_patch14_dinov2.lvd142m", pretrained=False, num_classes=len(class_names))

ckpt = torch.load(weight_path='deepfaune-vit_large_patch14_dinov2.lvd142m.v2.pt', map_location=device)

state_dict = ckpt['state_dict']

new_state_dict = {k.replace('base_model.', ''): v for k, v in state_dict.items()}

model.load_state_dict(new_state_dict)

but it fails with this error:
Error(s) in loading state_dict for VisionTransformer:
       size mismatch for head.weight: copying a param with shape torch.Size([30, 1024]) from checkpoint, the shape in current model is torch.Size([26, 1024]).
       size mismatch for head.bias: copying a param with shape torch.Size([30]) from checkpoint, the shape in current model is torch.Size([26]).
 

Are you using a different backbone for v2? I tried BACKBONE= '"vit_large_patch14_dinov2.lvd142m.v2" but that also doesn't work.

Simon Chamaillé-Jammes

15 October 2024 12:53pm

For the record now that this is here:

This error typically occurs when the wrong number of classes is given to timm.create_model. 
You should try to specify num_classes=30 manually (for this v1.2, number can change in future versions as we add new species).

Also, for issues please do reach out (as Jennifer did) by email, we will be much more responsive. More general questions/discussions can be asked here, I will reply asap. 

See full post
discussion

Building a biodiversity startup focused on getting landowners to use native plants

Colleen

23 September 2024 3:37pm

I'm fresh off my MEM/MBA degrees building a startup called EarthScape that is focused on assisting landowners in the switch from turf grass lawns (the largest USA irrigated...

8 2
See full post
discussion

AI and its environmental issues

Jennifer Lisa Insupp

26 September 2024 4:11pm

Hi all,I just wanted to get your insight on a topic that lingers on my mind: I was recently at a conference under the umbrella of UNESCO about AI and its influence on the planet...

1 2
Frank van der Most

26 September 2024 6:55pm

Great and difficult question Lisa! I think, yes, we are adding to the problem. We have been already when ML was introduced as a solution to big data problems, and perhaps we have been long before that with things that I am not aware of. My guess would be that the conservation sector is too small to have a big influence, but probably that can be guessed for many sectors. So, I don't believe we are off the hook.

Today at the acoustics webinar we had yet another discussion that ended up with how much we are trying to achieve with how few means. AI and ML allow us ( or hopefully will in the nearby future ) to do a lot of data and analytical work - mostly recognizing animals and their behaviour in sensor data - that without these means would be unthinkable within the resources that are available.

These are just guesses and believes. I would like to know how the damage weighs up against what we can do with it for nature's benefit, but I would not know where to start.

The argument seems similar to that related to questions like 'should we step in airplanes for conservation and climate conferences?', 'Driving cars around in the field?', and even 'use all those electronics and batteries?'

See full post
discussion

Conservation tech in Human Wildlife Conflict

Angelamercy Baltazary

12 September 2024 1:50pm

I am looking   for effective and low cost   conservation tech as a tool for Human Wildlife conflict mitigation. Anyperson with an idea kindly help.

11 0
Brett Muir

25 September 2024 4:56am

Odor based methods would be interesting. Provided they didn't need to be replenished too often.

We find varied stimulus prevents habituation.

See full post
discussion

AI & Illegal Wildlife Trade

Ahmed Junaid

9 September 2024 9:00am

How effective is the use of AI in monitoring and regulating the illegal wildlife trade? Are there any notable case studies and research examples available demonstrating its...

4 1
John Simeone

20 September 2024 5:08pm

Hi Ahmed - I'm involved in a couple different initiatives that are leveraging AI/ML to assist in monitoring and detecting potentially suspicious and illegal trade. Here are some papers that we've already published -- and we have several more in the works/under review currently. I can post them here in this thread once published: 

●Truszkowski, J. M., R. Maor, R. Bin Yousuf, S. Biswas, C. Chater, P. Gasson, S. McQueen, M. Norman, J. Saunders, J. Simeone, N. Ramakrishnan, A. Antonelli, and V. Deklerck. pre-print, forthcoming. A probabilistic approach to estimating timber harvest location. https://ecoevorxiv.org/repository/view/5059/

● Datta, D., N. Self, J. Simeone, A. Meadows, W. Outhwaite, N. Elmqvist, and N. Ramakrishnan. 2023. TimberSleuth: Visual Anomaly Detection with Human Feedback for Mitigating the Illegal Timber Trade. Information Visualization. https://doi.org/10.1177/14738716231157081

● Datta, D., S. Muthiah, J. Simeone, and A. Meadows. 2021. Scrutinizing Shipment Records to Thwart Illegal Timber Trade. Outlier Detection and Description Workshop, ACM SIGKDD 2021. https://oddworkshop.github.io/assets/papers/7.pdf

● Datta, D., M.R. Islam, N. Self, A. Meadows, J. Simeone, W. Outhwaite, C. Hin Keong, A. Smith, L. Walker, and N. Ramakrishnan. 2020. Detecting Suspicious Timber Trades. Proceedings from the Thirty-Second Conference on Innovative Applications of Artificial Intelligence (IAAI-20). NY, NY. https://doi.org/10.1609/aaai.v34i08.7032

● Mortier, Thomas, Jakub Truszkowski, Marigold Norman, Markus Boner, Bogdan Buliga, Caspar Chater, Henry Jennings, et al. “A Framework for Tracing Timber Following the Ukraine Invasion.” Nature Plants 10, no. 3 (March 2024): 390–401. https://doi.org/10.1038/s41477-024-01648-5.
(The researchers from World Forest ID who lead the above paper - together with Virginia Tech Computer Science, UW Remote Sensing and Isotope labs, and me are all now on a NSF grant to further this work together). 

Also, here are a couple of other recent relevant papers by other teams of researchers: 

● Gore, Meredith L., Emily Griffin, Bistra Dilkina, Aaron Ferber, Stanley E. Griffis, Burcu B. Keskin, and John Macdonald. “Advancing Interdisciplinary Science for Disrupting Wildlife Trafficking Networks.” Proceedings of the National Academy of Sciences 120, no. 10 (March 7, 2023): e2208268120. https://doi.org/10.1073/pnas.2208268120.

●Wing, Kate, and Benjamin Woodward. “Advancing Artificial Intelligence in Fisheries Requires Novel Cross-Sector Collaborations.” ICES Journal of Marine Science, August 28, 2024, fsae118. https://doi.org/10.1093/icesjms/fsae118.

See full post
article

🌍🎶 **Inspiring Moments at the African Bioacoustics Community Conference!** 🎶🌍

Consolata Gathoni Gitau

18 September 2024 3:27pm

The  African Bioacoustics Conference!

3 6
CathyNj

18 September 2024 5:37pm

This is very exciting! Thank you for sharing.
Grace Mchome

26 September 2024 1:08pm

Exciting! what indicators did you use to track land degradation and restoration using acoustics?
See full post
discussion

Camera trap function in heat waves

Karen Mager

29 August 2024 9:59pm

Camera trap functionality during heat waves:Hello my fellow camera trappers! Our cameras have suffered this summer during southern Oregon's heat waves. We have had a couple south-...

5 0
Akiba

10 September 2024 3:35am

Hi Karen. 

It's actually possible that it might not necessarily be the high heat that is causing problems with the batteries. Lithium should be quite durable for the temperatures experienced inside the enclosure. Instead, the problem could likely be condensation. In an enclosure, if the sun shines directly on a sealed enclosure, the enclosure can internally heat up, evaporating any moisture in it and raising the dew point. If the external temperature then cools, the internal enclosure air that comes into contact with the cooler walls of the enclosure can form condensation droplets. If you've ever left a glass jar in the sun, you'd see this phenomenon. 

The problem is that if any condensation gets on the internal camera circuit board, its possible that it can potentially conduct enough current to increase the discharge rate of the batteries or in some cases, cause the whole device to go down. We've seen this happen in some of the field devices we deploy and have since protected critical circuitry against condensation.   

To avoid this situation, I'd recommend trying to keep your camera traps out of direct sunlight and putting a lot of dessicant inside before sealing them. Not sure if this solves your problem, but hopefully it might help. 

Brett Muir

10 September 2024 11:04pm

This is good advice. Most lithium batteries should operate fine within a -10c to 60c temp range.

Personally I've only ever encountered issues from cold weather conditions.

Non rechargeable batteries tend to have even wider operating temperatures. 

LFP Rechargeable batteries also have better temp ranges.

Word of caution regarding dessicant packs is make sure they aren't touching PCBs. We've seen cases where the packs absorbed moisture but then became damp enough to short out components.

Could also be a bad batch of batteries. Make sure to pre charge them if using rechargeable batteries and measure them while they are connected to a load. Bad batteries will often measure fine voltage wise when not connected to anything but once a load is introduced their voltages drop significantly more than good batteries.

 

Rob Appleby

11 September 2024 4:55am

Hi Karen,

Yeah the discharge curves of lithium cells tend to be very stable for a long time and then drop sharply at the end. Akiba and Brett's advice below re: condensation prevention is another great recommendation and could well be the root of your problem. Let us know if you have any questions or if you do any testing etc., as this sort of information is gold!

All the best,

Rob

See full post
discussion

Global model for Livestock detection in airborne imagery - Data, Applications, and Needs

Ben Weinstein

12 August 2024 8:19pm

Hi all,I was at a AI for ecology working group a few weeks ago and was asked to look into an airborne model for detecting livestock to assist in land-management, agriculture and...

6 3
Elsa Carla De Grandi

4 September 2024 6:15pm

It looks like the website has not been updated since 2022 and indeed you need to register. If you go under 'Manage Account' you may be able to register. I tried to register but I got an error saying 'Login failed! Account is not active'.  I got an email saying "Your account must be approved before being activated. Once your account has been approved, you will be notified." So I am waiting for the account to be approved...

I'll keep you posted

Lars Holst Hansen

6 September 2024 12:46pm

Hi Ben!

Great initiative! 

A review of deep learning techniques for detecting animals in aerial and satellite images

https://www.sciencedirect.com/science/article/pii/S1569843224000864#b0475 

lists a number of data sets (incl. one published by you it seems...) 

Also, @dmorris keeps a list of Terrestrial wild animal images (aerial/drone):

https://lila.science/otherdatasets#images-terrestrial-animals-drone

which seem like it might be useful for you.

https://www.sciencedirect.com/science/article/pii/S1569843224000864#b0475https://www.sciencedirect.com/science/article/pii/S1569843224000864#b0475
Ismael V. Brack

9 September 2024 8:10pm

Hi @benweinstein !

  1. Surely a general detector might be very useful for detecting objects in aerial imagery! Maybe something similar to what MegaDetector does in camera trap images, generally detecting person, animal, and vehicle (and thus also empty photos). This could greatly improve semi-automated procedures. It could also serve as a first step for context-specific detectors or classifiers to be developed on top of this general one.
  2. There is also the WAID dataset that is readily available. Our research group could also provide images containing cows, sheep, and deer from South America.

Best

See full post
discussion

Spatiotemporal species distribution modeling

Navodita Mathur

6 August 2024 12:39am

Hi,I have a deep passion to contribute in the field of ecology, biodiversity and AI. I have a tech background and about a year of experience in AI. My current focus is on species...

2 0
Will Armstrong

6 August 2024 4:09pm

Hey, 

You're correct, traditionally species distribution models are spatial only, but here are a few ideas for how to incorporate time into your modelling: 

  1. Categorise the time dimension (E.G. Month/Season/Year, wet season/dry season etc) and use it as an input variable. 
  2. Incorporate phenological data (flowering time, migration period etc) into the input variables.
  3. Use time series data and use an appropriate model to take this into account such as LSTM.

A few things to bear in mind:

  1. We need the time of the X data to match that of the y data. This constrains the data that we can use. I'd recommend using Google Earth Engine to get the X data as you can query it for specific date ranges and they have a wide range of relevant datasets. 
  2. Whichever variables you use as input also have to be used for prediction.
  3. Any categorised temporal variables you choose have to be applicable to all species being modelled.

I hope this helps! 

-Will 

See full post
event

Catch up with The Variety Hour: August 2024

Alex Rood

20 August 2024 12:58pm

This month, we hear about acoustic imaging sonar that reveals northern elephant seal behaviour, GPS tags that monitor the success of rehabilitated and reintroduced Andean condors, infrared cameras that detect thermal...

2
See full post