How do you manage your trail cam pics?
- Thread starter timbratlr
- Start date
yoderjac
Well-Known Member
There are other products like that. In my case, I have some tech skills so I use general purpose tools. Full resolution pictures are sent wirelessly back to my camp where they are collected on the computer. When I'm at camp, I move them to my laptop. When I get home I process them. I first delete any pictures that don't contain target species. I then have some VBA code in excel. It organizes and renames all the pictures with date/time stamp and location. It also creates a CSV file that has all of the data collected by the camera as well as the file path and name.
Next, I use a relational database. I'm using MS access, but one could just as easily use a free one like PostgreSql. There is code in this database that imports all the data. It presents a form that includes all the data as well as the picture. I use that form to add humanly observed data from the picture (Count, Species, Sex, and such).
The relational database is queried from time to time to estimate populations, sex ratios, and such to support QDM management decisions. These cameras run 24/7/365 from the same location year after year. We run a query that counts deer from Jan-Apr which gives us an estimate of survival. We do a second query during the month of Sep when bucks, fawn, and does are much easier to distinguish. We use this data to estimate recruitment and sex ratios.
The advantage of using general purpose tools like this is the flexibility. We often find we want to ask questions of the database that we had not envisioned previously. For example, one year we used picture counts across time to show how deer were reacting to hunting pressure. You could easily see how as pressure increased from archery through muzzleloader and into general firearm season. The total picture counts varied little from year to year, but the number of day vs night pictures was directly correlated with the pressure. This is something you can do with a relational database, ask previously unanticipated questions, that you can't do with Excel.
The downside of using general purpose tools like this is that you need some fairly advanced computer skills.
Next, I use a relational database. I'm using MS access, but one could just as easily use a free one like PostgreSql. There is code in this database that imports all the data. It presents a form that includes all the data as well as the picture. I use that form to add humanly observed data from the picture (Count, Species, Sex, and such).
The relational database is queried from time to time to estimate populations, sex ratios, and such to support QDM management decisions. These cameras run 24/7/365 from the same location year after year. We run a query that counts deer from Jan-Apr which gives us an estimate of survival. We do a second query during the month of Sep when bucks, fawn, and does are much easier to distinguish. We use this data to estimate recruitment and sex ratios.
The advantage of using general purpose tools like this is the flexibility. We often find we want to ask questions of the database that we had not envisioned previously. For example, one year we used picture counts across time to show how deer were reacting to hunting pressure. You could easily see how as pressure increased from archery through muzzleloader and into general firearm season. The total picture counts varied little from year to year, but the number of day vs night pictures was directly correlated with the pressure. This is something you can do with a relational database, ask previously unanticipated questions, that you can't do with Excel.
The downside of using general purpose tools like this is that you need some fairly advanced computer skills.
Thanks for the insight and I like what you’re doing. I’m an “IT” dude so I have a few skills and it sounds like you do to. i used FileList to pull metadata and create a csv file that I modify prior to uploading to my spreadsheet. I perform a good bit of analysis using weather and moon data but you’ve taken it a bit further. Have you thought about packaging and selling?
BenAllgood
Well-Known Member
Some of the camera companies have the software built in or available. Tactacam, for instance, lets you run reports to see statistical data for bucks, does, fawns, predators, etc by wind direction, temps, moon phase, sun position, time of day, barometric pressure, location, and user assigned tags (like target bucks). It's come a long way.
yoderjac
Well-Known Member
There are many other products that do this sort of thing. I've evaluated some of them. Most are canned. They provide a few different standard views of the data. They do have some user bas where those views suffice for folks. They are generally easy to use and don't require any real computer skills.
The reason I'm using the general purpose tools is that I can analyze and view the data in custom ways that I had never considered. This is too complicated for the average user. Folks with the skills to do this, can easily use the same tools as I am an do the same thing.
I would also say that most guys are not using game cameras for management purposes like me. Most are using them for hunting purposes. Folks using them for hunting are generally trying to figure out when and where to hunt. I've tried correlating pictures of mature bucks with all kinds of meta data including weather, moon, rut based on fetus measurements, and many more. While we can make some general statements about, deer in general, does with fawn, bucks vs does, and mature bucks, the reality is that when trying to hunt a mature buck, you are dealing with a unique animal.
I think the best we can do with game cameras for hunting is to build our confidence that we are hunting in an area that is being used by a mature buck. Beyond that, I don't see a picture management system really helping folks harvest mature bucks which is what the market really wants. So, I would feel like I would have to over hype what a picture management system could actually do to sell it.
Many companies want your data and will give away software like this for your data. As Ben says, many camera companies offer on-line systems where you can upload your phots where they can be organized and managed and correlated with the metadata and external factors.
If you are looking for a hole in the market to fill, I would say it would be accurate AI identification of target animals in camera photos. This is the long-pole in my process. It is looking at individual pictures, zooming in when necessary, cleaning them up with photoshop when necessary to identify bucks from does and antler configurations. This is especially difficult at night with black flash. There are Convolutional Neural Networks (CNNs) that can do a pretty good job with humanly taken photographs. I have not seen one of these effectively applied to the wide variety of picture qualities we get from game cameras.
I would love to be able to run a list of jpeg files into a program that would extract the subject meta data from the scene (what animals, what sex, what maturity, etc. )
I'm waiting for this as well. I had a Moultrie cell cam 2-3 years back. I used it for a season or so. It could ID some species and buck/doe. It wasn't amazingly accurate, but it's good to see some companies are looking to get this functionality into the marketplace. AI likely isn't cheap enough to run yet, but image analysis is likely easier than the LLM stuff that's hitting the commercial market now.
yoderjac
Well-Known Member
AI is very cheap to run. The expensive part is building and training the CNN. In niches like military where the time has been put in to training sets, the technology is as good as a human. You need system with a fairly large number of GPUs to run the training sets on. This may be prices for individuals, but chump change for companies. The real cost is the human capital required to get a good training set. Someone has to do what I've been doing, analyze game cam pics, and identify ground truth to the extent a human can. This needs to be done with millions of pictures from many cams in many different environments.
Once trained, it takes little HP to run AI. You can run individual pictures against an established AI system with very low horsepower.
yoderjac
Well-Known Member
How well does it perform on deer on the fringe or beyond the flash range where a human needs photoshop to clean up the picture to identify the animal as a deer? I think current AI CNNs would do well on game cam pics over a point source attractant. It is in field situations where deer are distant and in poor lighting conditions that I have not seen good performance yet.
yoderjac
Well-Known Member
That is pretty good, but do you get false positives and false negatives, at about what rate? I'm using black flash with long-range PIRs. That means at night I often get eyes only. I need to clean up the picture to verify it is a deer, not a coyote or other animal. I can see that deer pretty easily without cleanup.
yoderjac
Well-Known Member
Keep us posted!
I agree and yes, I use for hunting and not deer management. Although, I do glean some information from my pics on my personal property but mostly use for hunting purposes. I’ve had some success with my analysis but again, whitetail deer can be very unpredictable once the rut starts. I agree the largest gap out there is identification beyond just doe or buck or whatever. Would like to see an animal identified and given a names that software can then identify every time it shows up. That’s what I do manually but have a process that somewhat eases that effort.
I really appreciate your insights and thanks for responding!
cahaba_hunter_20
New Member
Hi yoderjac, would you be willing to share your library of deer images you've collected? I'm working on a tool to estimate a deer's gross score from a photo and am looking to collect a training set exactly as you've described.
Last edited:
yoderjac
Well-Known Member
Actually, there are already tools for that. I evaluated one quite a few years ago. My library would not be a very good dataset for that. Tools like that work best with closer up images and multiple images of the same deer. To get a good dataset for that, one would want to follow the MSU deer labs established protocol for a baited survey where the bait is relatively close to the camera. My BEC Orion network has a long narrow PIR range. It is not uncommon for me to get pictures as far as 60 yards.
In addition, I'm hoping to retire at the end of the year and we are busy getting the current house ready to sell and building the retirement home, so I'll pass. Good luck with your tool!
What kinds of pics are you looking for? I may have some I can share from last year. How close does the deer have to be, etc?
cahaba_hunter_20
New Member
I'll take any picture you've got! Even if there is no deer in it at all.. So to be more specific about what I'm doing. Right now I'm training a classification model to look at a trail camera picture to decide if a) the picture has no buck in it b) has one buck or c) has 2+ bucks in it. I don't necessarily need any metadata or labels about what's in the picture bc I can label them myself if needed, but I'm aiming to collect 4-5k images if you have them and are willing to share.
Also, doesn't matter if the buck is up close or not.
cahaba_hunter_20
New Member
Just curious, what are the tools that are out there that do this already? I haven't come across any that are very good yet.
yoderjac
Well-Known Member
Your original post said you were trying to estimate a bucks gross score. It was quite a while ago, but there was a tool I evaluated. I think it might have been called trophy score or something like that. It was basically a measurement tool where you took measurements between ears and snout to calibrate antler measurements. You needed closer pics than I have for a tool like that.
In your most recent post you said you are trying to train an AI classification model. I think you will find that you need way more than 4-5K images to train a CNN classifier.
Having said that, good luck with your project! Some of the big AI companies have convolutional neural network classifiers that are general purpose and are as good or better than humans at identifying subjects in pictures. They have run literally millions if not 10s of millions of tagged images through them for training.
I think classifying deer from trail camera pictures is even more challenging. The data sets used to train the CNNs are generally scraped from the web. (Lots of legal work going on in this area). Those pictures are generally taken by humans and posted. Those are generally high quality images or folks would not post them. Lighting is generally reasonable, pictures are in focus, atmospheric conditions are generally good. Game camera pictures are taken when a PIR trips a camera in all kinds of conditions. Things like black flash are used to deal with camera avoidance. Weather conditions vary wildly from fog to sun glare and more.
The more variety in the picture conditions, the larger the training dataset needs to be.
My dataset has variation in tagging. My cameras run 24/7/365 for year, but we only extract and analyze a subset of data. For example, we use early Jan thru late March to estimate survival. We don't care about buck doe ratios for this, so while I tag the number of deer pretty well, if it is not obvious, I tag the sex as unknown. I don't take the time to zoom in or use tools to clean up pictures to identify the sex during this period. Many bucks have lost antlers anyway. The other period we analyze is the month of September. We use this dataset to establish buck:doe and doe:fawn ratios. During this period, I take much more care to identify sex, zooming in and using photoshop to clean up pics to identify sex. During this period, bucks have enough antlers to identify them as bucks and fawns are still easy to distinguish from does. For the rest of the year, I go back to being lazy tagging sex.
Are you a grad student working on a school project or is this a commercial endeavor?
Last edited: