Join us in conversation with Matthias Schmidt, CEO of Analyzer Pro. Matthias graduated in 2020 with a Master of Science in Physics from the Technical University of Vienna. He stands at the forefront of accident reconstruction, developing advanced software for collision reconstruction and speaking at international conferences like the European Association for Accident Research and Analysis EVU and the Austrian Association of Court Experts. Notably, as the CEO and one of the developers of Analyzer Pro, he drives advancements in the field. Beyond his CEO role, Matthias contributes significantly to industry journals like "VKU" and "Impact Magazin," influencing the evolution of accident reconstruction. He's also actively engaged as a member of scientific boards, an external lecturer at DEKRA Automobil-GmbH, and the organizer of crash test days in Murska Sobota, Slovenia. Join us for an enlightening conversation with Matthias Schmidt, a leader in accident reconstruction, and the driving force behind Analyzer Pro's innovation.
Originally aired on: December 21, 2023
Join us in conversation with Matthias Schmidt, CEO of Analyzer Pro. Matthias graduated in 2020 with a Master of Science in Physics from the Technical University of Vienna. He stands at the forefront of accident reconstruction, developing advanced software for collision reconstruction and speaking at international conferences like the European Association for Accident Research and Analysis EVU and the Austrian Association of Court Experts. Notably, as the CEO and one of the developers of Analyzer Pro, he drives advancements in the field. Beyond his CEO role, Matthias contributes significantly to industry journals like "VKU" and "Impact Magazin," influencing the evolution of accident reconstruction. He's also actively engaged as a member of scientific boards, an external lecturer at DEKRA Automobil-GmbH, and the organizer of crash test days in Murska Sobota, Slovenia. Join us for an enlightening conversation with Matthias Schmidt, a leader in accident reconstruction, and the driving force behind Analyzer Pro's innovation.
Originally aired on: December 21, 2023
00:00:30:27 - 00:00:50:07
Speaker
Hey, everybody, it’s Eugene, and welcome to Forensics talks. This is episode 94. And my guest today is Matthias Schmidt. We're going to be talking about collision investigations with Analyzer Pro. Now, as always, let us know where you are from in the comments section. So go ahead. Just type in there if you have any questions or anything like that.
00:00:50:09 - 00:01:09:13
Speaker
Today is in fact, prerecorded. However, we will get back to you on the comments. If there's any questions or anything like that, make sure you subscribe to the YouTube channel if you want to see more forensics talks and get notified. Whenever we put a new episode on, we usually try to put it on about a week before, and this gives people enough time to plan.
00:01:09:23 - 00:01:31:16
Speaker
Don't forget also that we have the Forensic Photography Symposium coming up this January 22nd to 25th, 2024. And basically, that is where we bring together a whole number of forensic photographers and just talking about a whole bunch of different things in different disciplines. So, it could be a collision investigation, it could be crime scene investigation, it could be for postmortem photography or whatever.
00:01:31:21 - 00:01:33:23
Speaker
And if you're interested in attending,
00:01:33:23 - 00:01:37:01
Speaker
it's the FBI. Let me just bring that up here.
00:01:37:01 - 00:01:48:03
Speaker
can register right here. Okay. So, it's just if you go to the website to Dash 3D dot com, you'll see it there on the main screen. You can just jump in any time. No problem. So, we're going to get going here right away.
00:01:48:03 - 00:02:01:27
Speaker
Not wasting any time. And I want to introduce the speaker. So let me bring up my notes here because I'm missing something, and I need to get my notes up here. Sorry about that. Just a second. So
00:02:01:27 - 00:02:10:14
Speaker
Schmidt is someone who is from the Technical University of Vienna. At least he graduated with a Bachelor of Science in Technical Physics back in 2015.
00:02:10:20 - 00:02:34:16
Speaker
He also went on to get his Master of Science in 2020, looking at physical energy, technology and metrology. He's worked in the area of collision investigation and accident reconstruction and analysis for about 20 years. But he's currently the CEO of Analyzer Pro and this is a software package for analyzing traffic accidents, and it has a number of interesting features.
00:02:34:22 - 00:02:51:06
Speaker
Now, collision investigation and accident reconstruction are super important. You know, it's something that unfortunately happens all too regularly, but we often have to determine who was at fault or what was the cause. And sometimes this also translates over to the criminal side because there can be
00:02:51:06 - 00:02:55:27
Speaker
criminal activity using vehicles. And in fact, you can use a vehicle as a weapon.
00:02:56:00 - 00:03:17:09
Speaker
And sometimes these are involved in homicide cases. So being able to analyze the physics of what is happening is super important. And so, we're going to bring Matthias in here right now. There is James. Yes, I do. A Hey, thanks for having me. And guess how you. Hey, I'm great. Thank you so much. Now we met. I'm not even sure.
00:03:17:09 - 00:03:36:23
Speaker
How did we meet? We met online or something like that. Yeah, I think so. Yeah, we met. We met online. We have been physically at the same place, I think at Rex, but we didn't speak. There had no chance. But I just, like, saw the stuff you're putting up in the industry, which is kind of unique, I would say.
00:03:36:25 - 00:04:08:13
Speaker
And therefore, I have to reach out to that guy, does cool stuff. And also from a technical perspective, it was quite interesting for me because I would say maybe five years back or so, 3D photogrammetry became quite popular in Europe and people loved it for doing like larger crash scenes. And then they always tried to do car scans and they failed miserably.
00:04:08:16 - 00:04:09:08
Speaker
And
00:04:09:08 - 00:04:11:16
Speaker
and then you had to started with like
00:04:11:16 - 00:04:33:29
Speaker
huge effort things, trying to record them. It never worked out well. And then when I saw what you could do with reconfigured for the cause, it was like, geez, that's the missing link. Yeah, yeah, yeah. Well, the was that was the was the Rex conference your first conference like big conference at Rex.
00:04:34:02 - 00:04:50:14
Speaker
It was my first like my first conference in the U.S. It was the biggest, I believe there is probably no bigger one than that one. Of course I've been at numerous EU conferences, but they are, I'd say, a fraction of the size.
00:04:50:14 - 00:04:56:28
Speaker
Yeah, so that but that was an experience. Yeah. Because I think they ran it again in like 2017 or something like that.
00:04:56:28 - 00:05:12:17
Speaker
But that was my first one too. And it was, well, like you said, it was so big that there were people there that I knew were there that I was trying to meet up with, that I had no chance. It just it was impossible. Was way too much going on. So, yeah, it, it was a very great lots of crash testing.
00:05:12:17 - 00:05:32:16
Speaker
Yeah, they did a really, really great job. Definitely if anyone is in this area watching and you haven't been to Rex, the next one, I'm not sure when it's going to be, but you definitely got to definitely got to get there. Yes. So, I want to talk to you about your background a little bit and kind of how you started your journey into the, you know, heading up the analyzer, pro
00:05:32:16 - 00:05:34:15
Speaker
development and everything else.
00:05:34:15 - 00:05:47:23
Speaker
Your background is in. You did a Bachelor of Science and so were you always sort of interested in did you know you were going to be going into accident reconstruction or were you looking at something else before you got into school? Well,
00:05:47:23 - 00:05:54:10
Speaker
essentially, I had no idea that this field even exists. Of course, if you think about it, it must exist somehow.
00:05:54:10 - 00:06:19:15
Speaker
But my first touch point was when I finished high school and I wanted to study physics and I come from a relatively small town called Salzburg. People might not know it because of Mozart. And I knew one guy who had studied physics there, and we had a common friend. They put us in touch and that was a
00:06:19:15 - 00:06:24:25
Speaker
60 plus year old guy and so we met.
00:06:24:27 - 00:06:52:28
Speaker
I asked him, hey, how is it to study physics? What are your career chances? He talked 5 minutes about studying physics and then about one hour about this super boring software for traffic accident analysis and which was and a probe back in the day to analyze it exists is probably. Yeah. Almost the same age as I am, and it started many years back as a is a toss program.
00:06:52:28 - 00:07:15:14
Speaker
The guy developed it for himself because he was working at a court in Austria. And if someone wanted to buy it, he could. If not, it was also found it was not his like it was not a commercial product in that sense. Well, anyways, this guy. No, we have a common friend and that common friend kind of pushed me in saying, hey, yeah, this guy wants to retire.
00:07:15:16 - 00:07:37:27
Speaker
Why not? Why not? But taking over his company. And at some point, I. I wrote him an email and said, hey, would you be interested? And he said, okay, sure. Let's do this. And I got my best friend on board. And back then we were 23. We had no idea how to run a business, no idea about traffic accident analysis and no idea how to code.
00:07:38:00 - 00:08:05:21
Speaker
Yeah, and we thought how hard could it be? Yeah. And what I think what thrilled me in the, in the in the beginning was actually this aspect of, of building something. And of course, it's much easier if you don't have to build it from scratch. And I knew there was a there were already customers. I knew the idea and the concept was good.
00:08:05:24 - 00:08:26:07
Speaker
I knew that people were afraid that this guy will retire and yeah, didn't, didn't know what to do. And yeah, so that was, that was the start, right? So, you were already so kind of like when you were finishing off school, you kind of had an idea that you were going to be getting in at some point, right?
00:08:26:07 - 00:08:54:08
Speaker
So yeah, you're kind of like planting the seeds and preparing everything for that. Yeah. So, I got started when, when I was still in my bachelor's and that dragged out my master's studies by quite a bit. But yeah, so that was, that was clear pretty early on. Okay. And so, because of what you're doing, I mean, a lot of it is physics based, but then there's also the, the programing side, the development side.
00:08:54:08 - 00:09:32:17
Speaker
So, in school where you were, you're working on both at the same time. Hardly. we had a few coding classes. Like when you study physics, you, you have that. But I wouldn't say it was a, was a particularly talented programmer. I wouldn't say I'm, I'm one today. But the good thing is my co-founder and, and I, we're still, we're still working together, and we have kind of established this dual leadership where he does the difficult coding part and I do everything around that.
00:09:32:19 - 00:09:39:23
Speaker
We have both studied physics, so we both have the background, which is super important, I feel, in that field.
00:09:39:25 - 00:10:09:00
Speaker
And yeah, so that's our nice split of work and okay, that worked out well. Yeah. So, when you so when you took over the software back then, in what state was it in when you first picked it up. well, well, so as I said, the guy who did it initially did it as his own project.
00:10:09:02 - 00:10:54:14
Speaker
meaning it worked in a way, if you did exactly what he thought you were supposed to click and do, then it went smoothly. But as soon as you tried to do something else outside the beaten path. Yeah. Yeah. Okay. Like it crashed and so on. So we actually spent a lot of time in the first few years to just take old parts, rebuild them from scratch, because what I have to say, the concepts that were behind that were super good and well-thought and yeah, just the, the coding part wasn't always so smooth, but now I can say we, we have cleaned most, most of the parts.
00:10:54:17 - 00:11:12:28
Speaker
Okay. And I mean a lot of the I mean, I don't know how I imagine over the years some of the algorithms, I mean the basic physics don't change, but maybe the way that we approach them with the algorithms stuff, there may be little tweaks or little changes that maybe make them more accurate. And that sort of thing is that's something that.
00:11:13:01 - 00:11:42:21
Speaker
Was that a true statement, do you think? Yes. So, what we are doing is we are releasing every year an update off of our software. So right now, we just recently released version 24. And in this update, there is usually a mix between new stuff that we just think is cool and old stuff that we think needs improvement.
00:11:42:24 - 00:12:12:29
Speaker
because yeah, I'm, I'm happy to say that we have a quite active community of users and once I tell them a few times that they should bring in their ideas, they actually doing that. And there are a lot of good ideas out there of my, my backlog of ideas is, is pages long. and of course people often say, hey, I like this or that module of, of the program, but wouldn't it be helpful if X, Y, Z and then we're doing that.
00:12:13:00 - 00:12:40:09
Speaker
Are we trying to at least say, yeah, okay, content reworking. You mentioned that initially, you know, if somebody wanted to buy the program, they could. It wasn't really like a commercial product. So, when you took it over, was there a small customer base already? Luckily, yes. That made our life much easier. so, I would say it started becoming a commercial product a few years before I started.
00:12:40:12 - 00:13:25:11
Speaker
So maybe 2 to 4 years before that. So, there was a stable customer base which since then grew by a lot. But yeah, it was, it was good enough to at least have a small, small salary essentially for us because we had to live from something. And yeah, so I think it's such a difference if you launch a product from scratch or if you already have people and it creates so much more credibility, it, it gives you so much more feedback because yeah, it doesn't, it doesn't help if you sit in your very theoretical I'm a physicists tower and I develop what is smart no people need to tell you what they
00:13:25:11 - 00:13:44:29
Speaker
want and what they need. Yeah. And I imagine doing you get good feedback from your customer base. You have like good people that are helping out with that. Yeah, yeah, yeah. On all levels. Like, and I think every feedback is quite valuable. Even if people write, hey, I'm sorry, I don't find where this is and that's, that is a clear sign.
00:13:44:29 - 00:14:08:28
Speaker
Okay. We haven't been clear enough. It's, yeah. To, to the point where people just say, hey I have privately developed this calculation tool, wouldn't you want to, to build it in. And of course that's, that's also great. yeah. Okay. So how, how, how do you go about validating like some of the, you know, the algorithms and different things?
00:14:08:28 - 00:14:39:18
Speaker
You mean you write something, you know, maybe it's generally accepted that this is, this is the, the formula that you use for this or something like that. But is I mean, is validation of part of your process? It has to be, obviously. So, I remember or sometimes I get this question like, is this ISO certified? And I have to say, no, it's not because there is no ISO certification for a traffic accident analysis software against what would it be certified.
00:14:39:21 - 00:15:13:22
Speaker
So unfortunately, such a thing doesn't exist. So, you have to really go into the paper direction on one hand and so trying to get papers out there, we have, since we're talking about recs, there is one currently on the review about one of the crash tests of the wrecks. So that's one big thing doing papers. another thing is people are doing like we are used by some universities, and we have like master's students.
00:15:13:24 - 00:15:46:01
Speaker
So that's another thing. And what we have done both internally as well as in larger groups, is doing our own crash tests. So, there was a group that, like many years back in Europe that did a lot of crash tests for program verification, double checking results. And yeah, that's also that's also thing. And last but not least, we are trying to be very transparent with what we do.
00:15:46:01 - 00:16:10:19
Speaker
So analyzer is by no means a magical black box that will just yeah put out some result and say this is, this is the solution good luck at some we have like an impact report tool that gives you it gives you a list of these are the numbers, these are the formulas. How you get from A to B, and these are the results.
00:16:10:19 - 00:16:35:23
Speaker
So essentially, everybody who is capable of holding a calculator can double check. Yeah. And the combination of these things I would say does the trick so far. Yeah, that's an important point, especially if it's going to be used for any kind of trial or forensic investigation or whatever it might be, then yeah, the fact that you can drive these things or you can explain them, you know, fundamentally is super important as opposed to black box.
00:16:35:23 - 00:16:51:10
Speaker
You know, it just does it, it just happens. If you can't explain it, then they think that that's going to cause some difficulty at trial. Yeah, for sure. Let me ask you about the I mean, you've been in this for about ten years now, so the technology in the vehicles has been changing over the past ten years.
00:16:51:10 - 00:17:11:15
Speaker
And so, there's a lot more information with respect to the EDR or the crash data recorders and things like that. So, have you had to what have you had to adapt or what kind of changes have you seen that have been important to maybe not just your software, but anybody else's software too? Yeah, super big topic. Obviously.
00:17:11:18 - 00:17:38:05
Speaker
I believe that's the job of the forensic engineer will change in the next couple of years. More and more away from this classical mechanics towards data analytics and what we have done and what helps us a little bit is the structure of how analysis built up. So, everything is structured in modules and analyze it and we can super easily just add another one.
00:17:38:08 - 00:18:09:10
Speaker
And I think it started back then when people just wanted to do their driving tests with simple tools like even like phones, accelerometer accelerometers, and we just wrote the parser where you could read it in and then transfer it directly to a digital twin, if you will. and then in the last couple of years in Europe, the digital tachographs became mandatory, which are standardized.
00:18:09:10 - 00:18:36:07
Speaker
We have a standardized format which is always a dream for the for the programmer because nothing, nothing worse than non standardized stuff. and I would say nowadays it's no secret that the big topic is, is cd-rw. Yeah, that's, that's something we have been working on quite a bit recently, just sort of like the latest updates. So, a lot of things.
00:18:36:09 - 00:19:02:13
Speaker
yeah. And then of course we have other sources like video material. What I find important to say about all this data sources, they are important, they are great source of evidence, but they have to be handled with a lot of care. I've seen. And was that. Yeah. At the last few conferences there was a great presentation about a highway accident.
00:19:02:16 - 00:19:22:07
Speaker
and if you just had taken out taken to CD-R file and would have put it in and said that's, that's the real deal, it would have been just completely wrong because there was an offset of the steering wheel, and the guy would have gone in circles. So, so it was just completely, completely wrong in that case it was rather obvious.
00:19:22:07 - 00:19:53:12
Speaker
But I'm always trying to warn people, like when we host seminars and so on to say, Yep, use data, but think twice. Yeah, it's not it’s not the fast track. And honestly, it's even good for our industry. That's it's not the fast track because then otherwise anyone could do it and yeah. Are they're very well-established standards now or are you finding there is there a difference between you know European cars versus cars that are in North America or whatever.
00:19:53:12 - 00:20:17:03
Speaker
Like is all the data like just all different from everybody? And you have to accommodate for that. So yes and no. I'm super happy that we got in Europe this, this digital tachograph files because they are like not the not only the EU, it's like whole of continental Europe, including in the Can you explain that. What is that.
00:20:17:06 - 00:20:46:06
Speaker
So, it's essentially, it's actually three types of it's a device built into a any vehicle that is heavier than 3.5 tons with a few exceptions. And it's there basically three types of data recorded. One is social data that just half like has to do for like workers rights. And so that doesn't really interest us. And then the most common type of just records speeds the data in of one hertz.
00:20:46:08 - 00:21:16:06
Speaker
And then there is a bit of a rarer type that record speed data in even forwards. So great to get speed levels rough speed levels and to have an idea where to start searching these are standardized which I love. CD-R is a bit more difficult because we can of course we can't read in the direct all files since this is a proprietary format.
00:21:16:09 - 00:21:47:00
Speaker
But what we can do is we can read in the CSB files that you can export, and these are I wish they were a bit more standardized. Okay. with regards to countries and I know that in the EU it's mandatory to have it have CD-R into new newer calls I think from June 2024. but some manufacturers are doing it much longer.
00:21:47:00 - 00:22:15:16
Speaker
I just know Toyota is a very positive example of having tons of data readable for a long, long, long time. And yeah, so there are, there are differences, but the tendency is quite clear. It's, it's going to be more and more and yeah, yeah, I thought you meant you raise an interesting point that you know, the future of people that are doing all this, you know, reconstruction in a collision investigation.
00:22:15:18 - 00:22:39:00
Speaker
It's going to be more about analyzing the data as opposed to. Yeah, that's a very, a very interesting point. I mean, are you seeing that trend already? yeah, yeah, totally. There is. I mean, I wouldn't say there are no cases without data. Of course, if you've got older cars, you don't have that. But I would say like just a rough like feeling.
00:22:39:00 - 00:23:07:06
Speaker
It's almost 50% of the cases now and we at least have something. It's not that that the whole case is just a pure data analysis case, but some piece of evidence is there usually one car of the two or some CCTV or something like that. So yeah, I mean, it's a great opportunity also for the engineers out there to specialize in certain fields.
00:23:07:08 - 00:23:46:11
Speaker
I also believe it's going to maybe even pivot in a direction where you would have probably less the one size fits all expert, but then you might have one that is focusing on KDR or the next one focusing on video analysis, which is a trend we're actually starting to see. I feel. Yeah, yeah. Video for sure. Yeah. I want to ask you about video for sure because I was on the webinar that you did a month or so ago and you were showing some of the videos of work that you guys were doing in the Analyzer Pro Software, and I want you to talk about that a bit because I know video for me and
00:23:46:11 - 00:24:03:28
Speaker
for other people too. It's obvious it does in-car cameras and there's just cameras everywhere. So, there I can't see how there's going to be less of video. There's just going to be more and more video coming through. So, talk to me about what you guys have done in the end for the people that don't know what you've done in your software with the video.
00:24:04:01 - 00:24:24:09
Speaker
Yeah. Yeah. So first I have to say, this is a topic I really enjoy working on because that's a topic we can play around with a little bit. I know like the big buzzword, but it's, it's, it's interesting from a for us from a from a also from a research point of view and from a what's technically possible point of view.
00:24:24:12 - 00:25:12:00
Speaker
It's just super, super exciting stuff and it gets better by the minute and yeah, so what have we done? The first thing that we have done in that direction, which is I feel quiet, quite popular, is we have done a tool for static too, for getting speeds out of static cameras. So is typical CCTV scenario where you have a mounted camera somewhere and, and then you can you want to find out how fast the vehicles are and what people used to do or still are doing is they just count frames and then they look, this car has moved a pixel left or right or which works more or less, I guess.
00:25:12:00 - 00:25:41:09
Speaker
But it's a huge effort. And what we have done is we have trained a machine learning algorithm to detect not only vehicles, but also pedestrians and, and trucks and yeah, whatever. And it's essentially a two-step process. So, in a first step, you've got to rectify your video. So, you need a reference measurement or actually a several measurements.
00:25:41:11 - 00:26:15:18
Speaker
And with that we're creating an orthographic version of the video. So, we will see top down. We have also built in a tool that allows to remove barrel and cushion distortion. because usually let's say you have this typical camera that is mounted on a gas station, you will not have the data about if this is a white like which angle the camera is, it's just trying to capture as much as anyhow possible.
00:26:15:21 - 00:26:50:16
Speaker
Yeah, we can see it. We can see it here and on the right. and we came up with a solution where we say, okay, if you, if you are talking or if you have a video about the human made environment a.k.a a street, usually lines are paralleled there or in like in a rectangular shape and this algorithm tries to play around till the lines are more and more pellets.
00:26:50:16 - 00:27:19:06
Speaker
So, this is what you see here on the screenshot. This is pre a free removal of the of the distortion of the of the barrel distortion and yeah. And then you can just click a little button and it will, it will stretch to image in a way that's all these lines get parallel and with that we get quite neat auto graphic videos even without having any information about the camera.
00:27:19:08 - 00:27:55:26
Speaker
And yeah that's, that's quite helpful I feel. Yeah. So, once you have that so once you have your autograph, take video, you can just click play. And what it will do is it will, it will find all these vehicles. and then it just sees how far does the vehicle go with each timestamp. And from the rectification we have to in the translation pixel to distance and then we read frame by frame just the frame length and then we have time and distance and that equals to speed.
00:27:55:29 - 00:28:22:26
Speaker
So, you can see on the like in this little graph with the red line that it's a little bit zigzagging. So, it's not, it's not perfectly smooth, but then we usually put in a fit curve, which is, which is the blue one you can see here. And that gives us a pretty accurate pretty accurate. Yeah. Results. So of course, people ask then what is like, how accurate?
00:28:22:29 - 00:28:45:10
Speaker
And I'd say it's an error of I would say plus minus five in that case is 10% which is certainly not good enough to say hey, this is a car that's going to be 60 or 61 kilometers per power. But it's definitely good enough to see if it's going with 60 or 70. Yeah, sure. That's our goal with that.
00:28:45:13 - 00:29:16:26
Speaker
Right? Right. Yeah. That's great. Are you planning to do any more validation testing on that? That sounds really interesting. Yes. So, I previously mentioned that wrecks paper and that's part of it. and I, we are in discussions with someone who wants to work even more on that actually with two people. and so yeah, we have done our own tests and also presented them at the conference, but it's always good if more people do that.
00:29:16:29 - 00:29:38:14
Speaker
But for that one, I'm pretty confident that it holds what it promises. Yeah, it's pretty amazing. I think the just the automation is going to like for anyone who's out there who's counted frames in video, it's not a fun task you like. You're sitting there, you're counting, you're following people or following vehicles. So, this level of automation is amazing.
00:29:38:14 - 00:30:00:20
Speaker
Like that's really going to make it, you know, just so much easier for people. So, yeah, it'll be interesting to see the papers coming through and looking at the uncertainty and errors and I imagine it will change. Now there's something that I just I recall from the webinar you talked about a certain frame rate or something with respect to the video, something about ten hertz or ten frames per second.
00:30:00:23 - 00:30:26:00
Speaker
yeah, that's something different. So, I was about to say we, we have a second tool being almost ready. So that's about dashcam analysis in dashcams, you know, in some models of dashcams you have, you have the speeds like recorded and written down. Then, then it's easy. We also have a tool for reading all that, but that's not so fancy.
00:30:26:02 - 00:30:55:21
Speaker
But what do you do if you have just an optical flow? So, a video of a car driving, and you want to find out how fast is that car? And that's where we thought, okay, we could try something there. And what we did is we created tons of test drives, and we recorded a driving behavior connected to certain speeds.
00:30:55:21 - 00:31:25:14
Speaker
So, we know, okay, if, if the optical flow looks like this. So, pointing to the image coming closer to you that is connected to this or that speed and we have yeah. Collected ours and hours and hours of material and trained a machine learning algorithm with that. And the goal is, and I mean it's the proof of concept is there right now the training is still running.
00:31:25:14 - 00:32:03:21
Speaker
This always takes forever. But the goal is that you can just put in any video, and it will give you out of the optical flow to speed. So, coming back to your questions with the ten hertz, this can only work if the video you're putting in has the same general specifications like the material we trained it with. And since we can't assume all types of cameras and can train a model for each and every type of camera that's out there, we have to standardize something.
00:32:03:24 - 00:32:24:14
Speaker
And what we did is when a user puts his video in, it will just be converted to ten hits, which is, I would say, still a good enough frequency. Again, we're talking about we want to find out levels of speed, not plus minus one kilometer power. Right, right. yeah, and that's the story I got you about that.
00:32:24:16 - 00:32:52:24
Speaker
Yeah. This is super new. So, it is. It is part of version 24 about it. It's like, not fully available yet because if, if we put that out, we want to be really sure that it's that the results are good. And your problem with if you do machine learning you, you start the process, then it calculates for two weeks or so and then it says, I found an error and then you start again.
00:32:52:27 - 00:33:12:01
Speaker
Okay. Yeah, yeah. So, it takes a lot of time. I want to ask you about some of the other things that you have. I know you have some basic drawing tools that if somebody wants to create a sketch or something like that, they can put that in. You also have the ability to import, which I'm very familiar with, the 3D point cloud data, and that gets enmeshed and convert.
00:33:12:01 - 00:33:50:06
Speaker
It could be a roadway, could be whatever, so it could be drone data. So, you ingest different types of 3D data, which is really awesome. So that's a wonderful feature. The you find that there are a lot of people using things like drone data now scan data. yeah, yeah. yeah, yeah, yeah. It has become a standard I would say at least like I come from Austria, so I am most familiar of course with the German speaking world, Germany, Austria, Switzerland and there it's super common to really.
00:33:50:08 - 00:34:24:01
Speaker
Yeah. People use more 3D photogrammetry than laser scanners, mainly because it's much cheaper and it's caught. You don't really get paid if you extra if you buy a laser scanner. So, you have to the choice do it. Pick up a laser scanner $50,000 or to abide is 3D photogrammetry software $450. Yeah, yeah. And the results are good A pretty good I would even say there are some products out there I'm personally familiar with, I guess.
00:34:24:07 - 00:34:56:21
Speaker
But I, I know that there are other tools that do essentially the same thing and that's how we got started with that whole Yeah, 3D point cloud stuff. So, the good thing about that is once you have set up your workflow and it's not very difficult to set up a workflow in that direction, you can do, you're super quick, you're much quicker than going there and like measuring every single trace manually.
00:34:56:21 - 00:35:24:06
Speaker
That's, that's, yeah, it's like I know a guy who does, who does a 3D photogrammetry for every single case, not because he wants to go so much the extra mile, but because it's actually a benefit for him. He does want reference measurements. He flies with his drone up and down the scene and he's done. And then yeah, and that's also why I love record free data, to be honest.
00:35:24:06 - 00:35:48:16
Speaker
And I'm not paid to say that though. But just yesterday I worked with a guy from the US on a case who scanned a not a full scene, but like at least the part of a scene with the free country and that that just works super smoothly, you know? Exactly. Okay. These are the marks on the ground.
00:35:48:16 - 00:36:10:20
Speaker
I have this there. I can compare it to my images. I can rectify the images. Then I know where the final position of my car was when it was there. At that specific point. Yeah. It just increases the quality of work by ideally even decreasing the time you need. Yeah, I think that's the key point there. It's yeah, exactly.
00:36:10:20 - 00:36:30:11
Speaker
It's reducing the amount of effort that you have to put in, you know, trying to get that. Well, there's no other instrument I mean between photogrammetry and even the laser scanners, there's just no other instrument that can give you that much detail in just such little time. And especially the drone. I mean, a drone is roadways are very well suited for photogrammetry.
00:36:30:11 - 00:36:55:29
Speaker
They have enough texture. They have all everything that's required. So, you know, less than 10 minutes flight time, you've got your roadway. You've got, you know, 300 meters or 400 meters of roadway. You've got all the details and everything else. So, it makes a ton of sense. So, yeah, it's great that you've got that integrated. You tell me a little bit about some of the features regarding the kinematics and the driving dynamics and things like that in the software.
00:36:56:01 - 00:37:21:09
Speaker
So, analyzer was born as a kinematics tool in like in the very beginning, because if you break it down, what is really, really important as an accident. It's not important to have a I mean, of course it's important, but it's only a part of the of the game to do a collision analysis. What's really because we know that a crash happened it's quite obvious.
00:37:21:12 - 00:37:52:22
Speaker
But what we really want to know is how it happened and why it happened and how it could have been avoided. And of course, you can do all of this in a dynamic simulation, but kinematics is just so much faster and more efficient, and probably that has a little bit to do with the history of analyzes. So, in the court where it was originally developed for what we do, life negotiations.
00:37:52:24 - 00:38:20:10
Speaker
So, you go there which the case person says, I was doing this and that, and then you have to calculate life and say, okay, this is possible, this is not possible. And then Person B does the same thing and so on. And for that you just need to be you just need to be super fast. And, and that is where Analyzer has its roots and where I think we're pretty much unbeaten on the market.
00:38:20:10 - 00:38:44:05
Speaker
And but of course then over the time, all these things like collision analysis driving dynamics came in, but I would say the analyzer really, really, really shines. Is kinematics just doing like quick analysis? Okay, we have this and that behavior, but how could he see that person when he was there and how was the situation? DAY okay.
00:38:44:06 - 00:39:03:08
Speaker
How could have been avoided? That's things you do with a few clicks in analyzer and yeah but did I understand correctly then like so for example somebody is at trial or whatever is it's simple enough that you know somebody says well what if this scenario, while the car wasn't traveling at that speed, starting speed, maybe it started at whatever, 20 kilometers more.
00:39:03:08 - 00:39:25:19
Speaker
Yeah. And they could just run simulation. Yeah. Just you just change that one number, and you see it and you the new version of it. That's pretty cool. It's. It's also. Thanks. It's also like you can set up in in less you can set up in a file I think up to 16 vehicles. And that's not because we're thinking that you have like multi crashes six vehicles at once.
00:39:25:21 - 00:39:46:20
Speaker
But the idea is to have different versions of an accident like placed on the on the civil vehicle. Some saying witness A says this and that. So, we have this version and be equal one and two. But then witness B says something that is this and that. So, when you at court you can just switch on version and then someone says, but no, that's not my version.
00:39:46:20 - 00:40:10:17
Speaker
You switch on a vehicle three and four and show, No, that's the alternative version. And at the end of the day, our job is not finding out what's legally what's the legal right thing to do. That's for the for the lawyers or the judges. But we just have to put out technical possibilities and saying, okay, this is possible what you're saying, no, this is not possible.
00:40:10:20 - 00:40:37:12
Speaker
And yeah, you can just do that super fast in that context and even life if necessary. You have some you got some case studies on the website that I think are kind of interesting. But when you were talking about being a trial, have you ever been asked to help support people at trial or have you ever had to if you ever had to go to trial because of the software or somebody asking questions?
00:40:37:14 - 00:41:00:28
Speaker
So, no, I personally never had to go to trial because well, out of several reasons. First of all, if you want to testify in Austria at court, you need to be you need to do a state exam that you are that certifies you. And if not, I mean, theoretically, it would work. But it's not a common, common thing to do.
00:41:01:01 - 00:41:27:24
Speaker
And we have we have good experts. They can ask me for advice, but they don't need me to show up. Their second thing is, at the end of the day, of course, the expert is responsible for the results he's delivering. I can only I can only deliver the tools for doing that. But in the end the of the person is they expect it himself is responsible.
00:41:27:26 - 00:41:52:21
Speaker
But what I have done actually quite often and yeah is I'm helping experts on their cases like I think even this this week I had two cases I've been working on two cases where people say, I'm a bit stuck here and there, what would you think about it? And then I usually like e to solve the case for them or work it through together in an online meeting.
00:41:52:23 - 00:42:12:08
Speaker
Yeah, I would say I have. For a person who has never been life at court, I have a lot of experience. Yeah, Yeah, right, right. Yeah, no problem. Yeah, it's good. That's interesting. I mean, looking at the case studies here, just offline. But yeah, there's some interesting stuff they're looking at. There’re some different applications. Hey, there's something else.
00:42:12:08 - 00:42:35:09
Speaker
You're. I didn't. I didn't think about it, but you have something called Analyzer Smart. What's that about? yeah. So that roots a little bit in that whole idea of we are life at court, and we have to do, like, quick calculations. So that's it. A small smartphone app for Android and, and Apple where you have a few basic modules of analyzer.
00:42:35:11 - 00:43:17:28
Speaker
So, for example you have this like you have a total distance and you know, okay, he crashed here, we've approximated this speed and this was where he said he reacted And now you want to know how fast was he was driving, how long was braking distance? So like kind of simple, but yet somehow sophisticated kinematic tools designed with the idea to do them life at court or with also what's also a thing that at least to my knowledge, is only happening in Austria is that you go to the location of the accident with the judge and the witnesses, and then people testify and say, this and that.
00:43:17:28 - 00:43:41:29
Speaker
Then you are supposed to calculate them. Yeah. So that stuff, it's, it's a small it's, it's a small, small, neat tool for. Yeah. For exactly these purposes. That was it. But was that something that was prompted because people asked or something or is it's just an idea you had where because I mean are you finding that people are using their mobile devices more and more.
00:43:42:01 - 00:44:05:28
Speaker
You know that's actually we did that many years ago. That was actually started the idea came from the EU Austria. So, if is the European Association of Traffic Accident Experts and they have a country group and if you Austria came to me and said, hey, we would be interested, could you, could you do that? And I said, Sure.
00:44:06:00 - 00:44:32:25
Speaker
And because the formulas were there, the logic was there, Yeah, we just had to learn how to coach on the phones. Okay. It's interesting. The yeah, I mean, I, you know, with the record through the app and obviously people using their phones more and more for different as different tools, I'm just wondering if you know if there's plans to do other things with your app in the future or something like that or if there's other things we could add to it?
00:44:32:27 - 00:44:57:24
Speaker
I don't know. I think I think for doing full scale traffic and analysis on a phone, the screen is just too small. At least I can't think of a way to how you could do that conveniently, but well, never say never. Yeah, you never know. Well, I'm just wandering like in. Feels like somebody maybe, for example, police who actually go and they're at the scene or something.
00:44:57:24 - 00:45:15:07
Speaker
If they're if there are other things they could do with the phone. I don't know. I'm just throwing obviously we were considering we were actually considering at one point doing some kind of sketching tool that at the end of the day we felt like that's probably not worth the effort. And there are other there are already other tools, and there's no point in doing the same thing twice.
00:45:15:09 - 00:45:41:11
Speaker
Yeah, I'm just I'm curious about your the customer base because obviously, like you said, you sort of had a heavy concentration in the German speaking countries and but of course you're throughout Europe and you know, as you expand or whatever, I'm wondering if are you finding that there are different requirements or people asking or emphasizing different things in different regions?
00:45:41:13 - 00:45:51:10
Speaker
Quite a bit, I would say so short answer, yes.
00:45:51:13 - 00:46:21:15
Speaker
I think there is there are very different levels of education in different countries regarding that field. So, you have everything between I'm an engineer, I'm specific, I've specifically studied that field of work and I have taken the state exam to I'm just someone who testifies at court, but I don't know much what's going on. And I don't I don't say that to blame someone.
00:46:21:15 - 00:47:05:18
Speaker
It's the legal system that leads to a certain type of industry. And with that, different demands come. So the more the more you go in this over your engineers directions, the more it's about numbers, formulas, graphs, facts and yeah, and other people probably prefer more like this very visual approach where you may you say, okay, I want to, I want to show maybe my idea about the accident.
00:47:05:19 - 00:47:47:14
Speaker
I want to show a visualize what, what, what people told me how it looked like. given our backgrounds, I can quite openly say that we come from the engineering direction, and that's what we are. Do you don't or we are not a jack of all trades, and we don't want to become one. It's we are very clearly designed and made for the peoples, for the people who like get a solid physics based, mathematics-based approach on accidents.
00:47:47:17 - 00:48:15:04
Speaker
Yeah. Okay. But they exist. They exist in, in so far, I've met them in every country. Yeah, right. You in terms of like training, you know, obviously having a strong user base and a very knowledgeable user base is important. So, do you offer like your own sort of training? Do you have people that do the training for you or are there I know you have webinars; you've done some things like you do know your website; I think you have something.
00:48:15:06 - 00:48:41:11
Speaker
Yeah, it's a seminar, so you host like seminars. So how do you, how do you teach people about Analyzer Pro? What are the different ways you can do that. So well main source is of course the webinars, but also, we are trying to put out a lot of video content because not everybody has the time to sit down several evenings a week or like two straight to three straight days.
00:48:41:13 - 00:49:17:22
Speaker
So, what we are doing is we're putting out videos, we have like video tutorials, we have a series called Analyzer Shorts and A and yeah, all of all of all of these things. Then we do our life seminars like real, real life seminars, which I personally like the most, because then you are and then you can just like meet with people and talk with them and the breaks and so on and but at least since COVID, the online, the online seminars became more and more popular.
00:49:17:22 - 00:49:58:03
Speaker
And I feel after COVID, even if people could travel, there is there is a sentiment to rather do something online and I don't know, journeying for several hundred kilometers to just meet. So yeah. So, with that said, these are two different approaches. I would say our main problem, if you will, we've, we've reaching new people and is that that people of course need to make an effort to learn the software.
00:49:58:05 - 00:50:38:29
Speaker
We are trying to make it as easy as possible, but it's still it's not Microsoft Word we are talking about. It's a software for engineers and we are trying to lower that barrier. S as much as any help possible. But yeah so for that the Seminoles are quite good but also like usually when people say, hey, I want to try out your software and there is no upcoming seminar, I just give them video recordings of previous ones because for me what it what counts is the people that use analyzer can actually profit from it and yeah, and use it properly and that it actually helps them because otherwise my, my work is pointless.
00:50:39:01 - 00:50:56:09
Speaker
Yeah. Yeah. Well, you know different people learn differently and sometimes you know online is great like for software, like things that you don't have to sort of physically handle equipment or something like that. But at the same time, you know how easy it is to get distracted, right? If you're at work or something like that, you're trying to learn videos.
00:50:56:09 - 00:51:14:08
Speaker
So, in those cases, the in-person, you know, being away from the office and being, you know, the phone's not going to ring while you're in the office or something like that can be really helpful for you to learn as well. So, I think the in-person thing is really, really good. Totally, totally. I, I really like it most. Like, first of all, you get to visit a new place.
00:51:14:08 - 00:51:42:02
Speaker
That's awesome. It's so much less exhausting. I don't know how you're doing, but I find two of online course more exhausting than one day of a real-life course. And yeah, I don't really know why, but yeah, you just it just gets so much better feeling of okay, where is that. That other person right now have I completely lost that person or is that, am I too slow, Am I too fast?
00:51:42:04 - 00:52:05:26
Speaker
All these things are so incredibly hard to find out online. So, I really, really like real-life meetings. And of course, like we have there is one seminar in Austria that's like our traditional release seminar in November. And by now I believe 50% of the people are coming there because they know that we are going to a nice like tavern in the evening.
00:52:05:26 - 00:52:34:11
Speaker
And yeah, and it one guy is always bringing a guitar and so you can't do that online. Yeah, yeah, yeah of course of course. Yeah. Those are, those are important moments for sure. Help to build a community and well, just meeting different people. You learn from different people, right? So different users. And yesterday I had a user tell me something about the record through the app that I didn't know, so I was like, hey, I learned something so often.
00:52:34:13 - 00:52:57:10
Speaker
Yeah, that's great. I want to ask you about the future of Analyze approach and just sort of looking forward. You mentioned AI, you mentioned, you know, some of the things that you're working on with video and things like that but what is what is sort of on the short-term horizon for you? Like what do you see as the big opportunity and what do you see as more of the long-term opportunity with Analyzer Pro?
00:52:57:13 - 00:53:29:09
Speaker
Yeah, good question. so, as I said earlier, we have this our number one key strength in kinematics, and we have our number two new key strength in data analysis. And since data analysis is becoming more and more important, we are going to heavily focus on that. I would say this is, this is an evergreen. It's, this will be around for the next ten years plus.
00:53:29:11 - 00:53:54:27
Speaker
And by every year there will be new sources of data and new things to do. And at some point, the autonomous cars are going to come. And so, I would say this is, this is probably our main, main focus from personal interest. I, I love dealing with AI, even though I think I can still learn so much more about it.
00:53:54:27 - 00:54:41:04
Speaker
It's just a huge field. yeah, I'd say this is, this is probably something we're going to do more about. yeah, yeah. And then of what we want to improve on is I would say our main and only big a weak point is that we don't do multi-party simulation. so no, no crazy flying people. And even though it's sometimes more so than accurate physics, I still feel that this is necessary.
00:54:41:07 - 00:55:08:14
Speaker
And this is, this is the only this the only thing that I feel we, we really should like, improve our current existing product if not talking about the future. Look, so pedestrian collisions and that it seems like it's a complex. It is. I'm not saying that we can do pedestrian collisions. There are other there are other methods. Just they have I've made a video about that.
00:55:08:17 - 00:55:41:16
Speaker
There are other methods, but still I feel it's sometimes appreciated. If you see that dummy like figure flying through the air and yeah, so that's, that's, that's something I want to I want to work on next. Very complex stuff, especially if you want to do it good. Okay, I have a question. Well, just in terms of people if they're watching this video and they want to get, you know, or they're listening to the audio podcast and they want to get a hold of you or something like that, obviously there's the website and I've been popping that up.
00:55:41:16 - 00:56:02:00
Speaker
I haven't put up the contact. Let me just do that there. But if you go to it's just going to be analyzer dot 80 and if you head over there you can you just get the website and there's obviously a on the contact form or is this the here you get all the people here who are part of the team and you got emails and phone numbers or you got a whole bunch of people here.
00:56:02:00 - 00:56:22:26
Speaker
So yeah, yeah, actually I didn't go here before, but I recognize a couple of names already, so that's interesting. Great. And of course, you're on LinkedIn, you're on social media as well. I hope you don't mind. I'm just showing this is your public LinkedIn thing. So, if anybody wants to get a hold of, they have questions about analyzer, they can also find you on LinkedIn.
00:56:22:26 - 00:56:48:18
Speaker
Yeah, yeah, yeah, absolutely. Reach out to me. I usually reply quite quickly on all types of messages, so write me email writing a LinkedIn. you will get through usually. Yeah. Right. Well, listen, Matthias, thank you so much for your time. I really appreciate it. I love what you guys are doing. It's great. And yeah, it's been great just recently meeting you and a great working relationship.
00:56:48:18 - 00:57:07:05
Speaker
You've always been really helpful and everything, so I hope that continues. I hope we can do a bit more going forward and I'll definitely be paying attention to the video stuff that you guys are working on. I think that's super important, super helpful, a great innovation. So, keep up the great work. Likewise, I can return to flowers.
00:57:07:05 - 00:57:26:27
Speaker
I also think you're doing a great job. And also, I really like that someone is doing like a bit like a modern approach to that field of work with the podcast and so on. That's yeah, that just has been missing a feel. And so yeah, but I think you're seeing, you're seeing the success and it's very well-deserved.
00:57:26:29 - 00:57:35:24
Speaker
So, thank you. Yeah, thank you. I appreciate it and I enjoy doing it. I love meeting new people and I've been learning a ton just from speaking to people like you. So
00:57:35:24 - 00:57:37:12
Speaker
Perfect. Alright.
00:57:37:12 - 00:57:38:21
Speaker
All right, folks, that's it.
00:57:38:21 - 00:57:42:18
Speaker
That does it for this episode. You know, we're coming to the close on the year.
00:57:42:18 - 00:57:45:03
Speaker
Christmas is just around the corner. And
00:57:45:03 - 00:57:54:17
Speaker
if you are celebrating Christmas, I want to wish you all Merry Christmas. Have a great holiday. It's important to spend time with your family and friends, eat well, drink responsibly, as they say.
00:57:54:17 - 00:57:56:07
Speaker
And you know we are going to be back.
00:57:56:11 - 00:58:07:19
Speaker
I'm not sure if we're going to have another one next week, just before the new year, but we will be back in the new year, and I hope to see you soon. Hey, everyone, have a very, very happy Thursday and we'll see you soon. Bye.