Hi Roland,

This is really amazing, great to hear about your set-up! I'm just wondering what the overall cost was to set up this system? Just thinking in terms of setting up something similar in other parks and what they should expect with regard to price. Would also be great to hear about the overall effort, e.g., hours/team members required. It would be great to have this act as a blueprint for other organizations/research stations wishing to deploy a similar system within their respective national parks/areas/etc.! 

Hi Rolland,

Interested too, but why did you choose SigFox (a private network) rather than LoRa (open network)? 

Sigfox currently has some financial troubles that, don't know what it will become in the long term.

Has anyone tried running the MegaDetector model through an optimizer like Amazon SageMaker Neo? It can reduce the overall memory footprint and possibly speed up inference on devices like Raspberry Pi and Jetson.

Jamie,

I have been using a 12-card Lexar Professional Workflow system for several years to process images from 24 trail cameras.  The system has 4 bays, with 3 card readers per bay.  I wrote software to automatically download and rename the images using metadata read from the info banner burned into the images by the camera.  (Alternatively, the software can use the Exif metadata attached to the image.)  The software is described in a recent Wildlife Society Bulletin: 

TWS Journals

Camera trap time-lapse recordings are useful for understanding animal activity patterns, but they can generate large amounts of data that require time-intensive processing and storage. Machine learni...

The Wildlife Society

The software is available on github:  

Don't be put off by "time-lapse recordings" - it works on still images also.

Regards,

Mike Hilton

Have a look at @tessa_rhinehart 's TechTutors on scaling up acoustic surveys. This is addressed at 11 min: 

The "hexadecapus" is the hardware, but transfer is automated by naming the SDCards prior to use and scripts take care of the transfer. The scripts may be written up in the audiomoth guide but the kitzes lab have this on github (although it may be mac specific) 

No - the trap was in their path and they just walked through it. I've now moved it to a place they can't go. The biggest threat to the moths is from pied currawongs. I schedule the trap so it shuts off at least two before sunrise to try to avoid them feasting on the larger insects.

At first I was finding wings below the screen in the morning when I put our units out. So I put a game camera on the units to see what was feeding and when. I found three bird species, likely 3 individuals, quickly found it to be a good bird feeder- Song Sparrow (most frequent), House Wren, and this Tufted Titmouse. I changed my units to turn off about 1.5 hours before dawn and that worked! Nearly all the moths left the scene before the birds came to visit. 

My most prized camera trap image - a hummingbird caught on camera!

Maybe a couple useful refs.

- we did find smaller detection areas in the wet season here

- guidelines for sample design here.

I've never had a problem with monkeys wrecking cameras, hyenas, elephants, bears yes.

cheers

Roland

Hi @laydent ,

I am just an occasional and non-professional user of a camera trap in Costa Rica. I have not experienced destruction of the camera trap by animals, so far. The trap never caught a monkey, but it did catch raccoons, an ocelot, pizotes and olingos. They all left the camera trap alone.

Rain would impact animal behavior obviously and possibly result in false positive triggering of the camera due to moving leaves and/or the rain itself. On the latter, one might try different sensitivity settings to see how your particular camera responds.

Camera traps are supposedly water tight ( and mine has been ), but in the long run seals may erode ( direct exposure to the sun may speed that up ). Also, I am wondering if water tight means rain-water tight, leaving the possibility of air humidity coming into the trap and cause corrosion.

Sorry, but I can not recommend anything to your third point.

I would agree that primates probably aren't your biggest concern, animal-destroyer wise. At least in the Malagasy rainforests, lemurs are highly arboreal so if you have trail cams down low, the lemurs won't come down that far. Neotropical primates are probably similar in that respect (at least compared to African/Asian primates such as baboons & macaques which are more terrestrial). The bigger issue than animals messing up cameras has been people stealing or interfering with them, so trying to put hem off-trail/hidden, locked onto large trees has been important. 

I second the Kays et al paper that @Rolandisimo mentioned above as a great reference & starting point! There is also this paper on camera trap study design, this one on camera trap placement bias and this one on strategic camera trap placement for evaluating different metrics. 

In terms of wet vs dry season, you'll want to make sure that water doesn't accumulate where you're putting up the cameras such that they would potentially become submerged in a flash flood or river-rising/overflowing situation, but you probably already know that :).  

There have been many papers that have looked at the effect of seasonality with camera trap data, but the majority of them are with regard to the actual species activity pattern differences across seasons rather than the effect of season on detection distance. You should consider what your species of interest's activity patterns are in different seasons and how this may impact detection probability. It may also be important in terms of strategically planning your sampling scheme. For example, the mouse & dwarf lemurs in Madagascar hibernate during the cold/dry season so it's not useful to sample for them during those months.   

I am a Camera Trapper hobbyist. I have gotten some nice otter videos. You can see them on my you tube channel pheekin trail camera videos. There is nothing on the channel but 15 second videos of animals that I have filmed. I have several otter videos on the channel that I think are pretty good. I found a spot where beaver, mink, and otters were moving between a medium size river and a man made wetland. I set my cameras on the trails. The secret was getting the camera really low on the trees near the trails. If you look at the otter videos the camera is pointed toward the river, just offset of the trail. When placed at 90 degrees to the trail the otters moved too fast for a good video. When offset but looking toward the river, the otters see the camera but most animals still proceeded down the trail toward the wetland. It helped that the river bank was steep and when they got to the top they triggered the camera at the same time they saw the camera. The camera was a true dark flash so they see the camera has and object but do not see the flash. One warning when cameras are set low on a tree the camera really has to be locked in tight. Raccoon and coyote will investigate the camera and some will knock it.  Coyotes knocked out  of position three of the 4 cameras that I set. Thankfully I got the otters before they knocked the cameras. Hope this helps 

Hi David,

Harold is correct. Cattle do like to use anything to scratch against. They are also pretty inquisitive.  

Cattle don't like shadows or flowing rubbish/ material. So attach a flag post near the camera or build a cage around it. Yet with a cage, cattle will most likely rub up against it. 

Never ending cycle of the scratching vs what you don't want moved

Hi Yvan, 

I dropped this into the AI for Conservation slack group as well, you got this reply this morning: 

Steph 

Hi Yvan,

If you find something which reaches your expectation and especially the ability to identify individual with plumage patterns I will be intereted.

In the same idea that Ultralytics, there is : 

Which could probably answer your first and second requirements.

And find here a great website showing all the IA-based camera software, you may find solutions or contact throw this list: