Doodler Posted January 18, 2021 Posted January 18, 2021 1 hour ago, hiflier said: Doodler, you and Believer57 would make the perfect team for something like you're describing. In a way I had pushed in the real world for taking environmental DNA sample from creeks and streams that drain large areas of watershed in order to test for any primate DNA that may be present. There is precedent for such an endeavor but again it could be costly unless the right lab would be available to do the testing. A cost break on the number of samples sent in would be good and some companies even provide the equipment for taking the samples at no cost. I've been waylaid by other projects both personal and other wise and this Covid thing has really put a damper on the whole idea but I hope to revive it when things get better. So ML/AI? Or e-DNA? Which do you think would be more viable at this stage of the game? ML/AI doesn't care what the data is, it just cares if it has enough of it. An acquaintance had a problem in her backyard with squirrels. She hates squirrels. So she wrote a very little bit of code, and plumbed it up to a camera feed, and had the AI text her whenever it detected a squirrel in her yard via the camera feed. It worked, literally 100% of the time. She had to add timers to stop texting her because she would get dozens and dozens of text messages with a picture of her yard, and a yellow box around the squirrel with a % guess as to how likely it was a squirrel, following the squirrel as it crossed her fence and made its way to the bird feeder. She did this in a day of tinkering and $50 worth of raspberry pi with a pi camera. A cat entered the frame, and the system ignored it because the likelihood of that thing being a squirrel was too low. You could easily apply this kind of algorithm to detecting humans in your yard as a better security system. You have to frame the problem statement in a way that gives the AI something to chew on. you can't just say "what's this thing?", because it has nothing to go on, you have to say "how close is this thing to that thing?" or "how much does this thing resemble that other thing?". Or maybe even, "how far from this thing is that other thing?" And you have to give it a whole bunch of known samples or it'll classify them into blobs of probability by itself. If you just said, "classify everything that moves in my yard" I think you could teach it to pull apart the frames of the video and box the squirrel into one category, the blowing leaves in another, the cat in another, birds in another, but it wouldn't know what any of these categories were, and not knowing what each are, you would have to decide which are interesting, and then teach it to ignore the cat or bird. If a dog walked through the picture, it would probably pop up a different category. At least this is what I infer from training on the subject. If you have DNA from a few thousand humans and known primates, I can only assume the AI would be able to be taught what was known human, what was primate and what was other. I've never done that myself (analyse DNA), I've only done basic stuff with predicting when an outage would occur to a computer system. If you just gave it thousands of DNA samples and ask it to group them together, that might work. Does someone have the decoded DNA from a bunch of samples handy for download? I have my DNA downloaded, or at least whatever 23andme gives you. I'm not sure it's 100% of my genome though, it could just be selected chunks. I'm 2% Neanderthal... which is fun. Perhaps someone should submit some of this DNA to 23andme. It would be pretty entertaining to see what they find, but use a fake name and address just incase the government shows up to kidnap you for being a bigfoot who can use a computer. I tell you what, if I worked for some genome company, like 23andme, I would be tinkering and hope not to get fired for it. I would grab the full DNA from every customer who agreed to release it for experimentation, and feed that into the AI as known-human, like they do with the Neanderthal variant testing, I think the end result would be "areas which diverge from normal human" maybe and then how much they diverge by %? From that you could pick out the most divergent? i.e. it wouldn't pop out a result of bigfoot, unless you have known bigfoot DNA. It surely could compare samples of unknown DNA and tell you how related those were or chunk them into categories. Like with the video images of squirrels saying 99% squirrel, this test result would say X% human, Y% primate, Z% like this other unknown sample. If the sample was badger DNA, that % would be really low. If the sample was bigfoot DNA, the % would be really high. In the end, you would have a spectrum of samples and how releated they were to oneanother, and could choose the extreme ends for further study. I'm sure there's way more to it than that, which is why I like the video idea, it's a much much smaller problem statement than DNA.
hiflier Posted January 18, 2021 Posted January 18, 2021 My novel has hit men working for corporate land developers killing BF's so the public would be aware that the creatures inhabit the area being developed. Government, aware of the BF's, doesn't know about the killings, except for the few greedy insiders running the illegal nationwide operation. What gets the story going is a hiker who found a dead one and stole a hand from it for proof. The rest....is history as they say. @Doodler There is a couple of good threads on everything you've said re:DNA, including a link in one of them to how science is categorizing and storing small fragments of unknown DNA. Whole cells acquired from fresh evidence, even from footprints in snow, could show a nucleus that contains the gene or genes responsible to either our Human brain, or that of a Great Ape such as a Gorilla or Chimpanzee. If you're interested, I can post links to those threads. There may be one that would be inaccessible without a Premium Menbership but there's also one or two here in the General Forum IIRC.
Doodler Posted January 18, 2021 Posted January 18, 2021 18 minutes ago, hiflier said: My novel has hit men working for corporate land developers killing BF's so the public would be aware that the creatures inhabit the area being developed. Government, aware of the BF's, doesn't know about the killings, except for the few greedy insiders running the illegal nationwide operation. What gets the story going is a hiker who found a dead one and stole a hand from it for proof. The rest....is history as they say. @Doodler There is a couple of good threads on everything you've said re:DNA, including a link in one of them to how science is categorizing and storing small fragments of unknown DNA. Whole cells acquired from fresh evidence, even from footprints in snow, could show a nucleus that contains the gene or genes responsible to either our Human brain, or that of a Great Ape such as a Gorilla or Chimpanzee. If you're interested, I can post links to those threads. There may be one that would be inaccessible without a Premium Menbership but there's also one or two here in the General Forum IIRC. yes, please. I'm interested in reading up on it. I've engaged my son (LOL, he's a 3rd year software engineering major and gets all squishy about AI) on the open source AI / machine learning toolkit side of things. I've used products one from a vendor named "Splunk" which is a big scale log aggregation engine, and used Python and R for programming, I've tickled Tensorflow for an engine on log analysis, and Spark for other big data analysis, but again, all text. If GE can use Tensorflow to pick out brain abnormalities, surely some hack civilian can use it to tell if there's a booger in a photo, or if DNA is different, eh
Arvedis Posted January 18, 2021 Posted January 18, 2021 (edited) 2 hours ago, Doodler said: ML/AI doesn't care what the data is, it just cares if it has enough of it. That is the purpose of ML, to be able to sift through that enormous stack of frame by frame images (thermal or natural lighting) with trained learning to recognize the environment - frame by frame. Think of the chunky bandwidth that would require by drone in difficult wireless areas. The military had to contract the design for such a fat network with their mini drones. The commercial market is not far away from that but costly right now. I think there is a very big difference between recognition of a squirrel, of which there are thousands of image references for an intelligent script to detect, and that of thermal images from 300+ feet interpreting an unknown landscape. A simple script will catch everything that moves with the wind. A good script will make perfect detections and tell you which variety of squirrel. So far, the only really successful AI/ML animal detection has been with koalas in their known habitat, and that was not done at night. That was done painstakingly close to the animals since koalas don't care they are being filmed. As far as alien hybrid experimentation, it sure looks like it. The mystery then becomes why the aliens felt it necessary to drop their abominations on earth. If they are such skilled space scientists then why not find some uninhabited planet for them. Instead they put them on earth where they are a toxic nuisance. Edited January 18, 2021 by Arvedis
wiiawiwb Posted January 18, 2021 Posted January 18, 2021 54 minutes ago, Arvedis said: Instead they put them on earth where they are a toxic nuisance. Toxic nuisance? What is the toxicity you allude to?
hiflier Posted January 18, 2021 Posted January 18, 2021 How about starting with this, Doodler: https://bigfootforums.com/topic/71415-sasquatch-vs-environmental-dna/
Doodler Posted January 18, 2021 Posted January 18, 2021 1 hour ago, Arvedis said: That is the purpose of ML, to be able to sift through that enormous stack of frame by frame images (thermal or natural lighting) with trained learning to recognize the environment - frame by frame. Think of the chunky bandwidth that would require by drone in difficult wireless areas. The military had to contract the design for such a fat network with their mini drones. The commercial market is not far away from that but costly right now. I think there is a very big difference between recognition of a squirrel, of which there are thousands of image references for an intelligent script to detect, and that of thermal images from 300+ feet interpreting an unknown landscape. A simple script will catch everything that moves with the wind. A good script will make perfect detections and tell you which variety of squirrel. So far, the only really successful AI/ML animal detection has been with koalas in their known habitat, and that was not done at night. That was done painstakingly close to the animals since koalas don't care they are being filmed. As far as alien hybrid experimentation, it sure looks like it. The mystery then becomes why the aliens felt it necessary to drop their abominations on earth. If they are such skilled space scientists then why not find some uninhabited planet for them. Instead they put them on earth where they are a toxic nuisance. Yeah, I hear you really have to know where the drop bears are. It's not going to be easy. I'm proposing identifying any heat source or vaguely human shaped heat sources, not necessarily targeting one individual beast shape. Offloading the processing to physical memory cards, instead of wireless transmitting will help. Offloading the navigation using GPS navigation on pre-programmed flight paths will help. Onboard real time decision making can take place, but maybe only dumb decision making like terrain avoidance and that's about it. I wouldn't want the drone to do the processing, in my fantasy setup. I'm under no illusions that this is easy or cheap. I'm seeing a home base, a team launching and retrieving drones, a team processing the memory cards, and putting extra batteries on charge. Make the machine at home, then turn the crank to make the sausage in the field, changing the recipe as you get feedback. Here's my fantasy weekend. 0) get FAA permission to fly drones? This will be interesting. What if they say no? 1) setup base camp somewhere handy 2) setup generator and gaming PCs with good GPUs for image processing, or use shore power 3) launch X number of drones covering a few miles of likely terrain chosen well ahead of time 4) rinse/repeat the drone launches all day/night long 5) as a drones return, handlers process the drone with new batteries and memory cards 6) operators processes the images, 1 to the AI, 1 to the backup disk 7) AI identifies interesting areas, operator programs the memory card and handlers launch drones in that direction rinse, repeat, wipe hands on pants 8) find Joe's lost dog or Timmy lost in the woods Be prepared to recover drones lost in the woods, I guess. I'll take a look at what the squirrel finder used for software.
SWWASAS Posted January 18, 2021 BFF Patron Posted January 18, 2021 0: They will say no if asked. 3: you cannot fly a drone out of sight of an observer without something similar to a IFR flight plan. . As of a new FAA ruling, you wiil have to have a drone license to fly one. That requires passing an FAA administered test. New drones require what is basically a transponder to identify the drone and its operator. I just bought a new 2020 transponder for my airplane and it was over $3000 A miniturized one would be even more expensive. Those without a transponder can only be flown in designated and preapproved areas. You can bet it will not be National Forest lands. If the government is involved in any sort of protection or coverup, drone operation in BF habitat would never be authorized. I was going to build a drone this winter and after reading the FAA rule making proposal, decided I wanted no part of it. It simply would not work for BF research. That not withstanding, someone could discretely do drone operations illegally. The problem with that is, a ranger seeing a drone would seek out the operator. With the forest fire hysteria, drones will be considered a fire risk. The way I read the new drone rules, it would be cheaper as well as legal to use a real airplane for BF research. Equip an aircraft with LIDAR and hidden BF trails and even BF could likely be detected. That may be a lot cheaper than a fleet of expensive drones. 1
Arvedis Posted January 18, 2021 Posted January 18, 2021 (edited) 1 hour ago, Doodler said: Yeah, I hear you really have to know where the drop bears are. It's not going to be easy. I'm proposing identifying any heat source or vaguely human shaped heat sources, not necessarily targeting one individual beast shape. Offloading the processing to physical memory cards, instead of wireless transmitting will help. Offloading the navigation using GPS navigation on pre-programmed flight paths will help. Onboard real time decision making can take place, but maybe only dumb decision making like terrain avoidance and that's about it. I wouldn't want the drone to do the processing, in my fantasy setup. I'm under no illusions that this is easy or cheap. I'm seeing a home base, a team launching and retrieving drones, a team processing the memory cards, and putting extra batteries on charge. Make the machine at home, then turn the crank to make the sausage in the field, changing the recipe as you get feedback. Here's my fantasy weekend. 0) get FAA permission to fly drones? This will be interesting. What if they say no? 1) setup base camp somewhere handy 2) setup generator and gaming PCs with good GPUs for image processing, or use shore power 3) launch X number of drones covering a few miles of likely terrain chosen well ahead of time 4) rinse/repeat the drone launches all day/night long 5) as a drones return, handlers process the drone with new batteries and memory cards 6) operators processes the images, 1 to the AI, 1 to the backup disk 7) AI identifies interesting areas, operator programs the memory card and handlers launch drones in that direction rinse, repeat, wipe hands on pants 8) find Joe's lost dog or Timmy lost in the woods Be prepared to recover drones lost in the woods, I guess. I'll take a look at what the squirrel finder used for software. 1 hour ago, SWWASAS said: 0: They will say no if asked. 3: you cannot fly a drone out of sight of an observer without something similar to a IFR flight plan. . As of a new FAA ruling, you wiil have to have a drone license to fly one. That requires passing an FAA administered test. New drones require what is basically a transponder to identify the drone and its operator. I just bought a new 2020 transponder for my airplane and it was over $3000 A miniaturized one would be even more expensive. Those without a transponder can only be flown in designated and preapproved areas. You can bet it will not be National Forest lands. If the government is involved in any sort of protection or coverup, drone operation in BF habitat would never be authorized. I was going to build a drone this winter and after reading the FAA rule making proposal, decided I wanted no part of it. It simply would not work for BF research. That not withstanding, someone could discretely do drone operations illegally. The problem with that is, a ranger seeing a drone would seek out the operator. With the forest fire hysteria, drones will be considered a fire risk. The way I read the new drone rules, it would be cheaper as well as legal to use a real airplane for BF research. Equip an aircraft with LIDAR and hidden BF trails and even BF could likely be detected. That may be a lot cheaper than a fleet of expensive drones. In theory, all looks reasonable until you get to LIDAR. This is not a device you plug in like a thermal. At least it isn't yet commercially. LIDAR is system with subcomponents that need to work perfectly in sync. And, you need to understand sensor tech. Small gaps or errors in sensor measurements can lead to errors in outputs. Has anyone seen amateur 3D topographic mapping? It comes equipped with some drone camera software and it is still a maturing area of tech. LIDAR is for pros seeking precise topo imaging. Without pro LIDAR expertise you won't generate quality results. Most drone and plane operators have no experience with lasers and high end 3D imaging software. But let's say that we are airborne and some LIDAR tech has joined the crew, unafraid of a far flung expedition with little to no stipend. Of course, this expedition is already 6 figures in the red with the time, tech, and resources needed to get this project underway but all the same, LIDAR is giving us data. It is the terrain modeling of the ground that is partially visible from the air, not underneath sparse or heavy vegetation. LIDAR uses infrared light from lasers to map terrain. The mapping pulses don't go through vegetation canopy, they go around it. Mapping animals, people, objects in terrain under vegetation is not possible, even with LIDAR. If we used an alternative method to get a topo map of the area it would have to be off grid. You'd have to do practice flights and something like this stands a good chance to get noticed, even in remote locales. Though, probably as long as you keep your stuff off social media you can make some progress with it before it gets noticed by the wrong people. I'm thinking Port Chatham as a test area, as long as we're fantasizing. 3 hours ago, wiiawiwb said: Toxic nuisance? What is the toxicity you allude to? Just my theory based on whatever presumed or anecdotal biologicals that have come up in reports and research. I didn't have the patience to photoshop this with a wilderness background. Edited January 19, 2021 by Arvedis
wiiawiwb Posted January 19, 2021 Posted January 19, 2021 When you say toxic are you referring to their behavior? Aggression?
Arvedis Posted January 19, 2021 Posted January 19, 2021 (edited) 28 minutes ago, wiiawiwb said: When you say toxic are you referring to their behavior? Aggression? I was thinking their biology, though that is interesting you applied it to human-like behavior. I wonder how they would react in a mindspeak conversation, "you know your behavior is toxic, right?" Edited January 19, 2021 by Arvedis
Incorrigible1 Posted January 19, 2021 Posted January 19, 2021 My eyes glaze over with the mention of mindspeak.
Arvedis Posted January 19, 2021 Posted January 19, 2021 2 hours ago, Incorrigible1 said: My eyes glaze over with the mention of mindspeak. I changed the text: 4 hours ago, Arvedis said: I wonder how they would react in a samurai chatter conversation... 4 hours ago, Arvedis said: I wonder how they would react in a rhythmic wood knocking conversation... 4 hours ago, Arvedis said: I wonder how they would react if a habituator exchanged twinkies for marbles and called it a conversation with a Bigfoot. 1
SWWASAS Posted January 19, 2021 BFF Patron Posted January 19, 2021 20 hours ago, Arvedis said: In theory, all looks reasonable until you get to LIDAR. This is not a device you plug in like a thermal. At least it isn't yet commercially. LIDAR is system with subcomponents that need to work perfectly in sync. And, you need to understand sensor tech. Small gaps or errors in sensor measurements can lead to errors in outputs. Has anyone seen amateur 3D topographic mapping? It comes equipped with some drone camera software and it is still a maturing area of tech. LIDAR is for pros seeking precise topo imaging. Without pro LIDAR expertise you won't generate quality results. Most drone and plane operators have no experience with lasers and high end 3D imaging software. But let's say that we are airborne and some LIDAR tech has joined the crew, unafraid of a far flung expedition with little to no stipend. Of course, this expedition is already 6 figures in the red with the time, tech, and resources needed to get this project underway but all the same, LIDAR is giving us data. It is the terrain modeling of the ground that is partially visible from the air, not underneath sparse or heavy vegetation. LIDAR uses infrared light from lasers to map terrain. The mapping pulses don't go through vegetation canopy, they go around it. Mapping animals, people, objects in terrain under vegetation is not possible, even with LIDAR. If we used an alternative method to get a topo map of the area it would have to be off grid. You'd have to do practice flights and something like this stands a good chance to get noticed, even in remote locales. Though, probably as long as you keep your stuff off social media you can make some progress with it before it gets noticed by the wrong people. I'm thinking Port Chatham as a test area, as long as we're fantasizing. Just my theory based on whatever presumed or anecdotal biologicals that have come up in reports and research. I didn't have the patience to photoshop this with a wilderness background. Lidar is already being used in drones. LIDAR can get through most tree canopies in the US because there are enough holes through the folliage that it can reach the ground as it flies along. Much of Western Oregon and Washington have been LIDAR scanned and anyone, including myself, can access it and use it to see what is under the trees. The State Department of Geology of both states have the data available to anyone that wants to use it. 1
Recommended Posts