 All right. Yeah. All right. Sorry, I had some, like, video issues in Francisco help sort those out very heroically. All right. Shake out. Okay. So this is when boats fly, accident reconstruction and blender. So did that go that way? Click. I have to be closer. Okay. There we go. I've been a long time blender user since, like, 1999, 1998. I do lots of things, but today it's, I don't know, no. Sorry about that. Weird. I love transitions. So sometimes, if it doesn't go, you maybe have to click a few times. Okay. Just keep clicking until the slide comes. Sure. Okay. No. Can I click a mouse instead? Actually, okay. Well, it works from this distance. Okay. People getting hit by cars, cars getting hit by cars, people getting hit by trucks, a car hitting a van and then a truck and then two more trucks, because that's what Americans do, and a flying boat, a boat hitting a boat. Okay. So all this is forensic animation. And forensic animation is a visual reconstruction of an accident or a crime scene. So in this line of work, accuracy is valued above all else. Recurations are pretty ugly, pretty basic as far as visuals, but the accuracy has to be spot on. So my workflow might include examining the testimony and the evidence, trying to understand the case, building the scene, like building the models and the environment, simulation and animation, and then examine the reconstruction and then iterate, try to make it more accurate, try to make it match testimony better. And so, as a starting point, I'll use visual evidence, sort of a case, I might get security camera footage, which sounds ideal, but it's almost always just terrible. It's low res, it's blurry, there's dirt on the lens, poor frame rate, poor angle. Because it makes sense, because if you got excellent security camera footage, they wouldn't need an animated reconstruction. On scene photos, it's fairly common, you might find things like skid marks, fluid trails, the final vehicle positions, or I might get nothing at all, which is, it sucks, but it happens. In that case, you have to rely on witness testimony and your own measurements, but problem is people are bad at estimating. If you give someone something to estimate, say, speed, distance, acceleration, deceleration, or reflexes, time, they will get all of it wrong. So you have to take some of that with a grain of salt, but it's sometimes all you have to work with. So, here we go, building the scene. So go and inspect the actual scene and the vehicles, measure and photograph everything. I'll take a couple hundred photographs and dozens of measurements. Google Maps is a big help for assembling larger scenes where you can't actually measure everything, and also for building in background detail. You can use pre-made models if it's a common enough vehicle, but go ahead and make sure you check every measurement that's accurate, and above all, keep it simple. It's not special effects. It's not beauty awards. It's just cold accuracy. So onto the simulation. Spoiler, it really doesn't involve me or Blender. Blender's physics simulations are excellent, but they're geared towards animation, they're geared towards special effects, not simulating a vehicle or collision between a vehicle and a person. So, and also, I don't have an advanced physics degree, which doesn't help. So I work with a physicist. He does his own simulation and I import his data and I animate the non-essential parts, say secondary motion, background animation, and so on. So to do that, I use Python. This is a very, very old script that I wrote in 2.4. The physicist works in a software called MS Mac 2D. It's a very, I guess, very battle-tested vehicle simulation software. It's essentially outputs an ASCII table of the vehicle positions, rotations, and steering angle each hundredth of a second. And the script just brings that in as IPO curves. So I'll run through a simple reconstruction example. This one was very nice in that it was low speed, just two common vehicles, very minor damage. And we had on-scene photos showing the final resting positions of the vehicles. So here's the on-screen photo. The driver of the Volvo, she pulled in front of the 18-wheeler. It was just a little tap. Only damage to her vehicle was the rear left quarter panel. It was a scrape across there, indicating that she actually brushed up against the truck as she passed by, and a broken taillight. Very minor damages, but she was claiming some massive injuries. So here is the output of MS Mac 2D. This is what the physicist did. This looks like a VHS recording. It is, actually. He's kind of old school. So it's just a little tap. It was about two, two and a half miles per hour, like maybe four kilometers per hour. And then I took that, I took the spreadsheet of the vehicle data, and I brought it into Blender, brought in my 3D models, and matched it up against the original animation from MS Mac 2D. It's just that small. So next we'll show what that actually looks like in 3D. So this is what they had to deal with previously, just a very basic user imagination kind of video. But of course with Blender, we can actually take a realistic viewpoint. Yeah, that's it. And the driver testified that he could really could not see the car pull out in front of them and stop as fast as possible. So having a full 3D recreation, I was able to bring a camera up to the driver's point of view and show what that looked like. And sure enough, with the tall hood, you can barely see her as she's pulling through. So all of that for a little tap. Yes, Americans love to sue people. We love it. It's what we do for a hobby. So the driver claimed major injuries, and I was working for the defense, so the insurance company. For them, it's actually cheaper to settle a case than to bring it to trial. They have a certain number amount that they know, OK, it's cheaper to do this rather than risk going to trial and risk some million-dollar judgment against them. It's cheaper to settle for X amount of money and just wash it away. So with all that came a challenge to my credibility. So the opposing attorney said, that's a nice cartoon you got there. And that was a fair statement. It's something he's probably never seen before, and I need to prove myself. So to review, I have no physics degree, but I'm showing a car crash. I have no computer science degree, yet I'm writing Python code. I just got a film degree, so people automatically think fiction and storytelling. And it doesn't help that my last project actually was a cartoon talking chicken who falls down in the end. Yeah. OK. Thanks, but that does not help in a legal case. Yeah. So to overcome those challenges, I was just fully transparent. I explained my workflow in a very detailed manner. I emphasized that I'm just doing the visualization, that the accuracy comes from the physicist, and I do the visualization. And the two do not meet. We talk to each other, but what he does, what I do, are completely separate. And also, that 3D skills are universal. If I can do an animated talking cartoon chicken, I have enough skills to accurately recreate a 3D scene of any type. So that worked out well. The next challenge, then, was blenders create a credibility. So you just downloaded this free software off the internet, and now you're doing Accident Reconstruction. Pretty much those exact words. And I try to think of a good comeback from that. But I was just like, yeah, that's what I did. So OK. There's an inherent bias there against free software, especially in the legal field, because everything costs a lot of money in the legal field, a lot of money. So precedent also is very important in law. They want to know, is it peer reviewed? Has it been used before in legal cases? And, yeah, all right. Not in the sense that they're looking for. There's no society of blender forensic animators or Sobfa for short. I just made that. There's no Sobfa gala. Sorry. But at the same time, though, it doesn't matter that there's no precedent for its use in legal cases. The math behind it is all solid. The software itself is peer reviewed by nature. So I explained what open source software is to them. And that they're free to hire an expert to examine blender source code, my source file, my source code, all of that. They're not going to do that, but they absolutely can, which is even more transparent than if I used, say, some other software with falling ceiling tiles. And so with that, then, went to a, so that played out well in the deposition, the first questioning of the whole process by the opposing attorney. And so then actually going closer to court was what's called a do-bear motion, which is a motion to exclude expert testimony. So for testimony to pass, it has to pass two tests. Reliability, technique can be tested, explained, peer reviewed and had a known error rate, et cetera. And relevancy, that the testimony is based on the facts of this particular case and that those techniques are properly applied. So with all that, actually entered the source code of my importer into evidence, yeah. And we passed the do-bear motion, which was opened up a lot of... That opened up a lot of doors, actually. This animation was admitted into evidence in federal court and the case settled. That's the goal, actually, of most of these cases. They want to settle and they worked. I got my foot in the door as far as forensic animation. So what I'll show next is a, what you all came to see. Oh, there we go. A complex reconstruction example. This one involves boats. This is a lot more difficult than that. Nothing is static. The ground beneath you is shifting and moving. There were no photos or video because, well, where are you gonna take a photo? And one boat actually flew. So the story is this work boat with the, it has a V-shaped hull and it has a steeply raked, can I say, steeply raked profile, hit this fishing boat and the fishing boat lost. It was a grandfather, father and son, on the boat. The father had some pretty severe injuries. The hull tore across the back of the scalp. They said they actually had to give him a facelift to stitch them back together and make them whole again. And lacerations across his back. And it was heavy damage to the fishing boat. And because of the shape of the work boat and how fast it was going, it was going about 40 knots, which is just flying, that work boat, let's see, there we go. The work boat was actually what's called on step, which is where it's almost doing the wheelie in the water as it's traveling down its path. So first thing was to examine, oh, sorry. First thing was to examine the boats, like 136 photos and of one, 119 of the other dozens of measurements, drawings, built a detailed cockpit. And some of the key findings from all of that were detailed damages. There we go. So being able to actually go and measure where the damages happened, how much deflection was in the sheet metal was really useful. Oh, okay. The angle of impact was one of the only things we had to go from. So on the left is the control console for the fishing boat and on the right is its engine. You can see the angle going across these matches and also you can see the blue paint on the engine. So we know that the hull crossed there at some point. So other key finding was, aren't you working? There we go. Shape of the work boat hull. We just sampled it every couple of inches and the physicists used that to create a profile. And with this, actually doing the calculations and the modeling isn't really, don't really simulate this with a computer. At least he's rather old school, he did it by hand because yeah, the other software simulation he uses would not be able to do something like this. So based on the relative heights of the vehicles, the vehicles vessels, the way that this one was up on step, almost doing a wheelie and the shape of the hull and the buoyancy of the fishing boat, he actually calculated how the work boat actually took off. So on the left actually was just a height and time sampling at each 10th of a second. So he found the speed was from the damages and the testimony and the trajectory calculated from the speed and the hull shape and then I modeled the vessels. And then to actually start animating, this is where the clicker, okay. This is where Blender really came in handy. It was critical because all we had was the angle and the damages. So I started by overlaying the two vessels onto one another where the damages actually happened and just started from our known data, the position and angle and then refined based on the physical evidence, the injuries, the damages and just try to make everything match as well as possible. Something like this, you have a lot less certainty than a car crash where you have a photo of the finished scene but we did absolutely the best we could. So here's the result. Boat comes along, they yielded as they were supposed to and the work boat just actually took off. It's kind of crazy. So let me show it to you in slow motion and try to explain all of the damages and why this happened as it did. Okay, so the V of the hull comes and makes its initial cut into the fishing boat, knocks down this chair, the father pushes the grandfather out the way and in the process, the hull comes across his back and his scalp, the corner of the hull scratches across the control console. The right motor hits the hull, hits something rigid behind it and kicks up and the left motor, the little fin at the bottom cuts across the hull of the fishing boat. It's a flying boat. You don't get your money back now, you have flying boats. So that's what we did. Another kind of key finding, actually putting ourselves in the driver's point of view, the captain's point of view, I should say. And because of the sort of battering rams he had in the front of his vehicle vessel, I keep saying, he couldn't see the victims until a little bit, just before it happened. So it was very, very kind of difficult situation there. So the results, yeah, we had a compelling reconstruction but with less certainty, it would have been more difficult to pass the doe bear motion. We cannot say with nearly as much certainty as the last one where it was just two and a half miles per hour on fixed ground, no sliding, no crazy stuff. But at the end of the day though, it was a very compelling view of what happened, especially from the driver's captain's point of view. And if you're the other attorney, you don't want to have to argue against something like this when your work boat scalped a father in front of his grandfather and son. So the case actually ended up settling out of court. And so with all the moving parts in this one and all the many, many variables we joked around that this animation was based on a true story or boats for short. Yeah, we're gonna make some boats. So how much time do I have left? Oh, excellent, okay. That's more than I thought. Okay, so back. So future progress. I would actually like to get into court. So far all the cases have settled out of court, which is what the attorneys want. Again, it's a lot of uncertainty going into court. You might get handed down a million dollar judgment against you. I would like to actually get that experience. It will be a lot more stressful, I'm sure actually going up on the stand, but it's something I'd like to do. I've done about six or seven of these cases. One of them actually settled the morning of the court. Did I show up in my suit and everything? I was like, all right, I'm ready, I'm ready. What, it's settled? Okay. I would be interested in doing criminal cases. So far they've all been personal injury. I did start a criminal case, but that one, the case was dismissed. That one was kind of interesting. I used, and then I'm using a little bit of a different software photo modeler to, this vehicle got shot up and of critical importance was how, where was the shooter? So I actually went to the vehicle and inserted wooden rods through all of the bullet holes and took a bunch of pictures and matched that up and 3D and figured out the vectors of the bullets. Something I'd like to do in Blender one day. I'd like to experiment with VR. I recently got an Oculus Rift and I don't think it would be too useful in the courtroom itself and that it's a very solitary kind of experience, but I think in terms of increasing the accuracy of these reconstructions and having a driver or victim put on the helmet and be able to look around and say, oh, well this was over there. I was going, I felt like we're going faster than this just slower than that and so on. This could be interesting, could be a good tool. And so since I do have the time I'd like to show you one more project. This one, well this is a guy getting hit by a truck. He was a hopper on the garbage truck there on the left and was sort of crossing the street and got hit by this 18 wheeler. So with this one we started from security camera footage. So our hero there, Jeremy. That's him running off to talk to the driver of the truck, just follow the orange circle. Truck moves back a little bit and fast forward and then for some reason he comes and runs back. Yeah, and the truck hits him. So this is all we had to go from. This isn't, you know, like I said, security camera footage is usually terrible. This was a half VGA resolution, four frames per second and from a poor angle. So first thing I did was speed measurement. The driver said he was going about 25 miles per hour. So I took the wheelbase of the actual truck, 226 inches, and then set up a scale and measured that the 18 wheeler only goes 166 inches in that quarter second. And because it's quite far away, almost perpendicular angle, I'll be getting ignore perspective. So we found out I was going actually closer to 40 miles per hour. When he said it was going 25, so we had a fairly good case. And so then I built the scene as with the others. You know, went to the actual location and visited the vehicles. And here is the result. So this is just the video footage in Blender. Once I built this all and matched the camera angles, I can actually look outside of the camera's view and really get a very wide viewpoint of it. Yeah, and so I can turn off the camera footage and even hide the gas pump so we can really see what happened. So I go from this terrible security camera footage to this really vibrant reconstruction. And so I'll show you a perspective view of it. Oh, maybe this one, can you hit play? Oh, this one doesn't work. Okay, well come see me afterwards. Another sort of, oh, back one. This one play. Okay, I just wasn't on autoplay. So we'll look at the security camera footage maybe hundreds of times, but one breakthrough from actually doing the reconstruction was this truck right here, right? It was not apparent until we looked at the driver's point of view that the driver of the truck really had nowhere else to go that it was crossing at the same time as the victim was getting into the street. So the victim is not even, not actually a danger until he crosses that white line. So just some of the interesting sort of breakthroughs you get from taking a different perspective on things, something you can only do when making a reconstruction. And so very last one, this is just, oh, oh. This one is just, click, thank you. This is just like a quick view of what it is. This is another reconstruction. This one involved some camera projection, but I don't have quite enough time to talk about all that. I just thought I would show the basics of this one, so. Yeah. A lot of people getting hit with various vehicles and things. All right, so that is all. Thank you for your time. Thank you.