Hi Lucy

As others have mentioned, camera trap temperature readouts are inaccurate, and you have the additional problem that the camera's temperature can rise 10C if the sun shines on it.

I would also agree with the suggestion of getting the moon phase data off the internet.

Do you need to do this for just one project?  And do you use the same camera make/model for every deployment?  Or at least a finite number of camera makes/models?  If the number of camera makes/models you need to worry about is finite, even if it's large, I wouldn't try to solve this for the general case, I would just hard-code the pixel ranges where the temperature/moon information appears in each camera model, so you can crop out the relevant pixels without any fancy processing.  From there it won't be trivial, exactly, but you won't need AI. 

You may need separate pixel ranges for night/day images for each camera; I've seen cameras that capture video with different aspect ratios at night/day (or, more specifically, different aspect ratios for with-flash and no-flash images).  If you need to determine whether an image is grayscale/color (i.e., flash/no-flash), I have a simple heuristic function for this that works pretty well.

Assuming you can manually define the relevant pixel ranges, which should just take a few minutes if it's less than a few dozen camera models, I would extract the first frame of each video to an image, then crop out the temperature/moon pixels.

Once you've cropped out the temperature/moon information, for the temperature, I would recommend using PyTesseract (an OCR library) to read the characters.  For the moon information... I would either have a small library of images for all the possible moon phases for each model, and match new images against those, or maybe - depending on the exact style they use - you could just, e.g., count the total number of white/dark pixels in that cropped moon image, and have a table that maps "percentage of white pixels" to a moon phase.  For all the cameras I've seen with a moon phase icon, this would work fine, and would be less work than a template matching approach.

FYI I recently wrote a function to do datetime extraction from camera trap images (it would work for video frames too), but there I was trying to handle the general case where I couldn't hard-code a pixel range.  That task was both easier and harder than what you're doing here: harder because I was trying to make it work for future, unknown cameras, but easier because datetimes are relatively predictable strings, so you know when you find one, compared to, e.g., moon phase icons.

In fact maybe - as others have suggested - extracting the moon phase from pixels is unnecessary if you can extract datetimes (either from pixels or from metadata, if your metadata is reliable).

camtrapR has a function that does what you want. i have not used it myself but it seems straightforward to use and it can run across directories of images:

https://jniedballa.github.io/camtrapR/reference/OCRdataFields.html

I can help you with that, please send us some of the videos and will build a pipeline for it 

aakash at thinkevolvelabs dot com 

- Aakash Gupta 

Green AI Products — DeCaTron

DeCaTron is an AI-based platform for assessing and simulating the biodiversity of forested regions

Think Evolve Consultancy

ffmpeg + TrapTragger (Camtraps + re-ID) and Whiskerbook(re-ID)?

 

TrapTagger About

Camera traps are a powerful tool for studying wildlife and are widely used by ecologists around the world. However, these motion-activated cameras can generate immense amounts of data, of which a large proportion is simply empty images generated by the wind and other disturbances. Traditionally, these thousands of images were manually sorted by the ecologists themselves in an error-prone and extremely time-consuming process – often taking longer than the camera-trap survey itself. This presented a massive obstacle to camera-trap-focused research. As such, we wanted to develop a way to make the process drastically shorter, and more accurate. In turn this would allow ecologists to run more camera trap surveys more often, resulting in more complete data, and more statistically robust results across greater and greater areas. Moreover, this would allow ecologists to spend their new-found time on the analysis of the data itself, and make the world a better place.

WildEye

Whiskerbook

Whiskerbook is a visual database of jaguar encounters and of individually catalogued animals. The library is maintained and used by biologists to collect and analyse jaguar encounter data to learn more about these amazing creatures.

Whiskerbook

Hi All,

I understand Okala are doing a lot of work with ML and cameratraps, you could try and contact them.

Okala - Biodiversity Insights

Where Technology meets Nature

Zamba Cloud can handle video!

Hey there - we can help you, please send us some of the videos and will build a pipeline for it 

aakash at thinkevolvelabs dot com 

- Aakash Gupta 

Green AI Products — DeCaTron

DeCaTron is an AI-based platform for assessing and simulating the biodiversity of forested regions

Think Evolve Consultancy

Hi, 

our software Deepfaune works with videos. It is trained for the European fauna and try to put all birds in a 'bird' class but it includes a detection stage so at least it can filter 'empty' images or videos. Performance on videos is not as great as it is on images, so I can't promise it will match your expectations. But the installation on windows is easy, so give it a try and see for yourself. It's a point&click software. You can use it in English, French, Spanish, Italian.

https://www.deepfaune.cnrs.fr

DeepFaune

Un projet collaboratif, des millions d’images et de l’intelligence artificielle pour développer un outil de reconnaissance automatique des espèces animales en France

DeepFaune

Great discussion! Pet (and other 'home') cams are an interesting option as @antonab mentioned. I've been testing one at home that physically tracks moving objects (and does a pretty good job of it), connects to my home network and can be live previewed, all for AUD69 (I bought it on special. Normal retail is AUD80): 

On the Wifi front, and a bit of a tangent, has anyone done any work using 'HaLow' (see below for example) as it seems like an interesting way to extend Wifi networks?

Cool thread!

I will be testing Reolink Wi-Fi cameras in combination with solar powered TP-Link long range Wi-Fi antennas/repeaters later this field season for monitoring arctic fox dens at our remote off grid site in Greenland. The long range Wi-Fi antennas are rather power hungry but with sufficient solar panel and battery capacity I am hopeful it will work. 
I am looking forward to explore the links and hints above after the field season. 
Cheers,

Hi Ben,

I would be interested to see if the Instant Detect 2.0 camera system might be useful for this.

The cameras can transmit thumbnails of the captured images using LoRa radio to a Base Station. You could then see all the captured images at this Base Station, as well as the camera's battery and memory information (device health). In addition, you could also change camera settings from the Base Station so you would not need to reclimb the trees to change from PIR sensitivity high to medium for instance.

The Instant Detect 2.0 cameras also have an external power port so a cable could be run to the ground to a DC 12V battery for long term power.

If you wanted to, you could also connect the Base Station to the Cloud using satellite connectivity, so that you can monitor the whole system remotely, but this will incur more cost and power usage of the Base Station.

I'd be keen to hear your thoughts,

Thanks,

Sam 

Hello to all, new to this group. This is very exciting technology. can it work for ID of individual animals? we are interested in Ai for identifying individual jaguars (spots) and andean Bears (face characteristics). Any recommendation? contact? thanks!

German

Yes. I think this is really the way to go!

Here is another metadata initiative to be aware of. OGC has been developing a standard for describing training datasets for AI/ML image recognition and labeling. The review phase is over and it will become a new standard in the next few weeks. We should consider its adoption when we develop our own training image collections.

OGC Seeks Public Comment on New Standard for Training Data for AI/ML Applications - Open Geospatial Consortium

OGC TrainingDML-AI Standard defines the model and encodings for standardizing any training data used to train, validate, and test Machine Learning models that involve location or time.

Open Geospatial Consortium

For anyone interested: the GBIF guide Best Practices for Managing and Publishing Camera Trap Data is still open for review and feedback until next week. More info can be found in their news post.

Best,

Max

Hi @MaddievdW have you considered using a 'tunnel' to help, well, funnel, small critters into a space with a camera that makes it a bit easier for detections? A few years back, we made a PVC pipe tunnel with a protected food lure that seemed to work well with even cheapy trail cameras (and in fact, we ended up having to block some of the IR illumination using a few layers of athletic tape over the LED array). Here's a rather blurry image of a bandicoot we got:

And another of an antechinus:

We weren't that concerned with image quality, as we were actually testing an RFID logger in the tunnel also, and simply wanted to know if we got critters with no tags and what species (if possible). So one thing I'd definitely consider if you go down this path is focal length of the camera. Here's a previous discussion on a similar idea: 

Comparisons: Close-up Lenses for Camera Traps | WILDLABS

Hey Guys! We're setting up a camera trap within a woodpile to see what mammals enter/exit the nesting box we've included there. For this purpose, we'd need a camera trap with a close focus lense as well as the ability to download the data from the camera via USB cable (since the camera won't be accessible once its installed in the wood pile). Bushnell NaureView (119740) does have a close up lense included but does not support downloading data via USB Bushnell TrophyCam (119874) does support downloading data via USB but does not have a close up lense included. Does anybody have a solution the solves both problems? What would help a lot would be if there was a attachable close up lense that can be bought seperatly, but I haven't found any yet. Thanks in advance! Nils PS: Using a wireless transfer (WLAN / GSM) of the images is not an option

wildlabsnet

I guess you could always 'calibrate' it to a certain extent by just using a longer section of pipe. Here's what our set up looked like, but again, we had the RFID logger, so a camera-only version would be a bit simpler - the camera and battery was in the screw top section on the left-hand-side of the tunnel image and each entrance had RFID antennas linked backed to a logger in the same compartment as the camera:

All the best for your research,

Rob

Addressing each of the questions/issues posed: 

Triggering Camera:   

If you are getting triggers, but empty frames, during known visits by these lickety-split animals, the issue is the trigger speed.  Looking at the Browning selection guide, for example, https://browningtrailcameras.zendesk.com/hc/en-us/article_attachments/12697703673243

I see that the Elite-HP5 models have a 100 ms advertised min trigger speed, which is slightly (50 ms) than the Dark Ops Pro DCL.  This is equivalent to 2 earlier frames (at 60 FPS video), which could be significant with fast moving targets.  

I have found, BTW, counterintuitively, that for Browning SpecOps and ReconForce models (Elite-HP5) that camera gets to first frame sooner when taking videos vs. when taking stills.  I don’t understand this completely, but it’s a thing.  

If you are not getting any triggers, then the PIR sensor is somehow missing the target.  Make sure you understand the “detection zones” supported by your camera.  These are not published, but can be determined with some patience and readily available “equipment” – see my post on “Trail Camera Detection Zones” at https://winterberrywildlife.ouroneacrefarm.com/2022/08/01/deep-tech-trail-camera-detection-zones/

Putting more than one camera at a site may also increase the probability that at least one triggers (and may improve lighting, see below)

If you’re consistently missing triggers, you may have to consider a non-PIR sensor.  Unfortunately, this removes you from the domain of commercial trail cameras.  Cognisys makes a number of “active” sensors based on “break beam” and (now) lidar for use with DSLR-based camera traps.   You would also have to come up with your own no-glow lighting source, and hack the DLSR camera to remove the (built-onto-the sensor) IR filter.  In our experience, these sets are 10x more expensive and time consuming vs. commercial trail camera sets, and are only justified by the potential for (a few) superior images. 

The species-specific triggers and sets mentioned on this thread seem like a better option. 

Avoiding daytime false triggers: All the commercial trail cameras I’m aware of have a single type of trigger sensor. It is based on a Passive InfraRed (PIR) sensor and Fresnel lens.  Apps, Weldon and McNutt cover this admirably in Peter Apps, John Weldon McNutt, “How camera traps work and how to work them,” African Journal of Ecology, 2018.

These sensors trigger on changes in certain areas of the thermal field – in practice a combination of a heat and motion in one or more detection zones. They are not decomposable.

Some cameras (e.g. Browning SpecOps, and maybe the Dark Ops Pro?) allow you to set hours of operation so that the camera only triggers at night, for example.  This would cause you to (for sure) miss “off hours” appearances by your target species, but would avoid daytime false triggers. 

No Glow Image Quality: The good news about longer wavelength “No-Glow” flashes is that animals are less sensitive to them. The bad news is that the CMOS image sensors used by cameras are also less sensitive to the longer IR. Less signal leads to lower quality images.  Others have mentioned adding supplemental no-glow illumination.  An easy way to do this would be to set up two cameras at each of your sites.  When they are both triggered, each will “see” twice as much illumination, and image quality will be improved.  Browning SpecOps models (at least) have dynamic exposure control on video which allows this scheme to work (with only a frame or two of washout) while the algorithm adjusts exposure). For an example of this effect, see opening porcupine sequence in our video at https://www.youtube.com/watch?v=itx7KnlxKS4

Hi Maddie,

This camera has a very quick reaction time. 

Reconyx HF2X HyperFire 2 Covert IR Camera

The HF2X is a state-of-the-art digital camera with passive infrared (PIR) motion detection and a high-output covert infrared nighttime illuminator—all within a secure, weather-resistant case. One of the best-performing, most reliable trail cameras on the market, it is backed by a 5-year manufacturer warranty.

Faunatech

I'm jupping into the discussion, with a similar objective. I'm looking for a thermal camera trap, (I know cacophony). it would be use to improve invasive speices monitoring especially for rats and feral cats.

Any idea?

Thanks

Hi @mguins , as @kimhendrikse mentioned resolution (and also brand) for thermal cameras can dictate a big jump in price. GroupGets has a budget Lepton (FS - short for 'factory second' I think) if you wanted to check one out: 

Teledyne FLIR LEPTON® FS

Lepton FS is a non-radiometric 160 x 120 resolution micro thermal camera module with reduced thermal sensitivity, reduced scene dynamic range, and up to 3% inoperable pixels but the lowest price ever for a factory Lepton. These units balance performance and price, enabling monitoring applications where radiometry is not required and pixel-level image information is less important than broad thermal data. With the lowest cost per pixel in the Lepton family, low power consumption, and simple integration, Lepton FS provides integrators an appropriate thermal capability for innovative thermal monitoring products in smart building automation, security, occupancy sensing, and more. Just as we were the first place to get FLIR Lepton seven years ago, we are bringing this new Lepton to you first at a highly compelling price to make your thermal imaging application more economically viable. To celebrate, all accessory boards below are 10% off excluding the new tCam-mini Wi-Fi board.   160 x 120 RESOLUTION NON-RADIOMETRIC MICRO THERMAL CAMERA COMMON INTERFACES AND FACTORY SUPPORT TO SHORTEN TIME TO MARKET Standard Lepton mechanical interface, electrical interface, and US-based Technical Services team. 160 x 120 thermal pixel resolution Low operating power – 140 mW typical and 650 mW during shutter event Small 11.8 x 12.7 x 7.2 mm package ACCEPTABLE PERFORMANCE AT DISCOUNTED PRICE Affordable, non-radiometric thermal imagery and data. -10 °C to 350 °C scene dynamic range Thermal sensitivity <75 mK  ≤3% inoperable pixels BUILD INNOVATIVE THERMAL MONITORING SOLUTIONS Build innovative thermal monitoring solutions appropriate for heat, security, and comfort monitoring applications. Home and building automation Heat and occupancy sensing Security and location monitoring   Specifications Overview Sensor technology Uncooled VOx microbolometer Spectral range Longwave infrared, 8 μm to 14 μm Array format 160 x 120, progressive scan Pixel size 12 μm Effective frame rate 8.7 Hz (commercial application exportable) Thermal sensitivity <75mK NEdt Operability Number of non-defective pixels shall be >97% Adjacent clusters, rows, columns may contain defective pixels that are not factory corrected and unable to be corrected. 3% of operability failures allowed. Temperature compensation Automatic. Output image independent of camera temperature. Non-Radiometric Performance While some of the Lepton FS units may output radiometric values per pixel, the Lepton FS units are not guaranteed against any radiometric accuracy and users whom wish to use Lepton FS products for radiometric applications do so at their own risk. Teledyne FLIR and GroupGets will not support questions pertaining to calibrating Lepton FS units. Non-uniformity corrections Integral Shutter Scene dynamic range High Gain Mode: -10 to 140 degrees C typical Low Gain Mode: -10 to 350 degrees C typical Image optimization Factory configured and fully automated FOV - horizontal 57° FOV - diagonal 71° Lens Type f/1.1 Output format User-selectable 14-bit, 8-bit (AGC applied), or 24-bit RGB (AGC and colorization applied) Solar protection Integral Electrical Input clock 25-MHz nominal, CMOS IO Voltage Levels Video data interface Video over SPI Control port CCI (I2C-like), CMOS IO Voltage Levels Input supply voltage (nominal) 2.8 V, 1.2 V, 2.5 V to 3.1 V IO Power dissipation (Typical, room temp) Nominally 150 mW (operating), 650 mW (during shutter event), 5 mW (standby) Mechanical Package dimensions without socket (w x l x h) 11.50 x 12.70 x 6.835 mm Weight 0.91 grams   Environmental Optimum operating temperature range -10C to +65C Non-operating temperature range -40 C to +80 C Shock 1500 G @ 0.4 ms   Part Numbers 500-0771-FS1 Documentation Lepton Google Group Lepton FS Datasheet Lepton Engineering Data Sheet rev 204 Lepton Software Interface Description Document (IDD) rev 303 Advanced Lepton Usage on Windows Getting Started with Lepton on Windows and PureThermal Basic Features Getting Started with the BeagleBone and Breakout Board V2.0 Getting Started with the Raspberry Pi and Breakout Board V2.0 Lepton with Radiometry Quick Start Guide Lepton vs Lepton 3-App Note  Mechanical FLIR Lepton 2.5 and 3.5 3D Model (step) IDD CAD data LEPTON 2.5, 80x60 IDD CAD data LEPTON 3.5, 160x120 IDD CAD data LEPTON 3.1R, 160x120 Software GroupGets GetThermal Viewer (Linux and macOS) FLIR Lepton User App for Windows v1.3.2 PureThermal Firmware FLIR Lepton SDK Development Hardware GroupGets - PureThermal Mini Pro JST-SR GroupGets - PureThermal 3 GroupGets - PureThermal Breakout Board GroupGets - PureThermal Micro Breakout Board GroupGets - PureThermal OpenMV FlyTron - DroneThermal v4 Teledyne FLIR - FLIR Lepton Breakout Board v2.0 danjuliodesigns LLC - tCam Mini v4 danjuliodesigns LLC - gCore OpenMV - FLIR Lepton Adapter Module for OpenMV Cam H7 Tips for integrating Lepton into a Housing Recommended Material PORON® Socket Stack Socket > Lepton > Foam > Window > Enclosure Maximum Force < 1Kg (more force will damage the shutter) The foam must be cut out so that it does not interfere with the mechanical shutter pins. Thickness Depends on the enclosure and spacing, do not exceed 1Kg of force. In the News CNX Software - $99 Lepton FS module cuts the cost of FLIR thermal cameras by half This Model: Lepton FS, 160×120 pixels, 57° with shutter - NOT Radiometric Important Information PureThermal 1/2/3/Mini/Pro with Radiometric FLIR Lepton 2.5, 3.5, 3.1R requires specialized software to show radiometry such as GetThermal. All international customers that are ordering any Thermal Imaging Camera will be required to fill out an End Use Statement document before shipment. Once you place your order we will email you with the document within a few business days. We can only ship FLIR Lepton to North America, Australia, Switzerland,  United Kingdom and the European Union Countries. Any orders with a FLIR Lepton with a delivery address other than those will be canceled and refunded. We are sorry for the inconvenience. Number of non-defective pixels shall be >97% Adjacent clusters, rows, columns may contain defective pixels that are not factory corrected and unable to be corrected. 3% of operability failures allowed. Temperature compensation Automatic. Output image independent of camera temperature. Lepton FS units are not guaranteed against any radiometric accuracy and users whom wish to use Lepton FS products for radiometric applications do so at their own risk. Teledyne FLIR and GroupGets will not support questions pertaining to calibrating Lepton FS units. RoHS Certificate of Compliance European Directive No 2015/863 Declaration of Compliance REACH Regulation (EC) No. 1907/2006 California Proposition 65 Declaration ECCN: CAM 6A993A HTSUS: 9027.50.4020 Country of Origin: USA Part Number: 500-0771-FS1

GroupGets

They also have a bunch of other Flir products and boards for interfacing with Leptons etc., so worth a browse of the shop. It could also be worth taking a look at Seek modules, some of which @Alasdair has experience with : (e.g. 

S314SPX Mosaic Core Starter Kit 320x240, 57HFOV, FF

They also have modules you can connect to a mobile phone: 

Compact for iOS

@TopBloke I'd be very keen to see your Lepton camera trap too! 

Cheers,

Rob