How Katy Schildmeyer & Sue Kagan Are Transforming Fit & Sustainability in Fashion | PI Apparel

I have the distinct honor of presenting to you the award winners for the 3DRC Grand Challenge in the startup category. And yesterday Dinesh presented on the academic side, and today Katie will take us through on the startup. And just a reminder, this was around how generative AI could be part of the product creation cycle. So congratulations, we're thrilled that you won and looking forward to it.

Okay. So first off, I want to thank the IEEE and the 3DRC for opening up a competition like this for innovators and people who really want to solve problems, not just for the industry, but for also every single person in the world. Because let's face it, we all wear clothes. And generative AI is something that really requires accuracy of information when applied to engineering. And that is exactly what fashion designers do every day. We are engineers first and foremost. So our foundational tools must absolutely be accurate for us. And I've been working in 3D since 2016, and it's come a long way. A lot of the programs have done an extremely good job with their physics engines on textiles. Again, the last presentation shows We still have some room for improvement. And what I'm going to present today will show exactly where there's room for improvement within the avatar. And if you think of us, we're the foundation and our clothes are the scaffolding that go around it. And every architecture plan needs specific information. And that's exactly why architecture does not use AI to create a building. So we're no different. And I represent Design Cycle Solutions, which was— it had many names up until now, but we arrived at Design Cycle Solutions. And this is the first part of a group called Apotheosis. Anyway, we've got 8 current rigging systems in the market right now. And the rig system is that bone structure that goes underneath the body and it's weighted in a certain way to allow it to fold and manipulate like flesh. But we have some key problems. Everybody knows the spine, the pelvis, the crotch, and the armpit. I'm sure we have all been in a room where we're discussing crotches, armpits, and posture. [SPEAKING CHINESE] And then the next slide, we're showing an example of what we see commonly in a program, where I might import a pattern and use the morphing tool for the in-program avatar. Now, as a new designer who doesn't know anything about 3D or fitting in 3D, they may use this tool unknowingly and make decisions with production that might cost in the long run, whether it's waste, whether it's low sales, or they gain a bad name and reputation related to fit. And then the second tool was one that was actually suggested to use from another 3D program that's very popular. And I, I won't say names just out of respect, but this program shows the body has adjusted in different ways from the breast, chest, waist, and hip. And that's in relation to the undercarriage of an SMPL rig structure. There's not enough spines in that body to create fluidity in the morphology. Yeah? Now this is a video to show what the morphology on that figure would be if I was adjusting the bone joints as I should. So keep an eye on the buttocks region and the hip region and see how it changes over time. But look at the up and down of it, okay? That's it. That's going to be an important piece. It doesn't really change in volume. It just kind of hovers there, right, and moves up and down. Now, that's important because in this image, in the red, I have 4 different body types, right? No, but no, you wouldn't think that those are from the same body, right? Now, this was one person that was scanned with 4 different mobile body scanner apps. And the results are very different. Now this has to do with the backend of how those programs work, but also that undercarriage of the rig. So if they're using the SMPL rig, which is over to the right, that rig structure only has 4 joints through the spine, which would then cause different shapes when they're moving that spine up and down to meet the height. And then we stacked those body scans and each one is in a different color, but you can see that the posture is slightly different, the height is very different, the foot plant, which is very important, is very different, and therefore causes problems if you're using a body scanner app to define exactly what your fit is. So are— how are you doing due diligence on that fit becomes the question, right? And then this is a few more avatars for you. These are avatars from many gaming engines. So it can be Cinema— or not Cinema, but Character Creator 4 or Daz 3D. And I do have a very popular video on YouTube that I guess a lot of people watch about how to measure a Daz 3D figure, but I don't give the tricks of actually how it fails. I just give how to measure it. So I see a lot of really mistake-driven avatars as a relation to that video, admittedly. So I could be part of that problem. But in these avatars, they mostly carry lordosis or swayback as their posture. So if we're using that as a foundation of posture in the rig structure, we're probably going to miss how people actually stand. So I want everybody to just take a moment right now, check where your shoulders are at, check where your spine is at, check where your neck's at, maybe feel back here for how it bends. Um, and you'll see exactly what I mean here. Uh, so this is a video from a fellow peer in the industry in London. He works in Savile Row. And he had a woman come in for a suit jacket for her end of life. So probably the last garment she'll buy that's tailored. And this is what he had to do to fit her.

Coat for an 84-year-old lady client. She is super round-backed, head forward, and stooping. All three figurations in one. The coat will not sit on the dummy. She's got a massive, massive back balance. That's how much I have to push the collar on the dummy, or where I have to set it so the balance looks correct on the garment. Huge back balance, massive. Praise the Lord, hallelujah! These type of figures don't walk in every day, but that's the trade I'm in.

Ciao! This is enough. So that type of structure is definitely one that is rare for us to work with, but we have older people in our community and we should be able to fit for them, and yet we don't. I personally have had to alter 3 generations of elderly people's clothes in my family. So I know that it's a problem and the rig structure in 3D does not serve a purpose for that, that market right now. In terms of the discovery process, what we did was we took the Apotheosis avatar that we hand blend and rig in a certain way. And then we moved it into a shoulder position that is much more normal, more forward related to technology. And we started to see more smoothing in the back. The collar fit correctly in relation to the rotation of the neck slope and the clavicle and shoulder at the high point shoulder. Uh, the other one, we all know the middle one, and then the last one was that form at the very beginning of the presentation that had the breasts very high up and the cutting through the body, um, which of course is probably the worst fit, yet it was a product that was referred to us by a company that is supposed to serve the industry. So we know that when we went back to those companies, In some cases, they aren't doing their own due diligence on the avatars. Now, how does it impact us if we're using rig systems that were made for the gaming engine? Because really, all of those rig systems were made for the imaginary space. Nothing was made for the tangible space. And in that, we did our own surveys where we went out and talked to people. We asked them, With pictures, can you point out where things are fitting poorly for you, right? And they would say, things are short here. They're not, they're not wide enough. They're not long enough through the inseam of the leg. Even people that were 5'4" were saying, things are noticeably getting shorter. Now if we make short products, it's very hard to sell them to anybody but short people. So that's, that's one thing that we do need to keep in mind is it's always easier for the consumer to shorten by an inch versus us having thousands of pieces of inventory that we can't sell because it's too short by an inch. So we need to be really cognizant of how tall our avatars are and what our actual brand spec is in relation to height. Not just the width. And in that, 30%— 38% said that they were unhappy with that upper body fit, which would relate to the fact that the avatars right now have a spine that starts here and goes here, and then it goes right into the neck. Very few of them will have anything more than that. And if you hunch forward and bend, you kind of bend right here, so you need something. And then in the 3D pipeline and the cycles, the rig systems failed to deliver any kind of sustainability related to due diligence, like the due diligence process. They can't tell you how successful that fit market is or how they've tested it amongst other people with patterns and back and forth in between the reality the pattern, and the virtuality cycle. So that's an important question to ask when you're looking for something. And then increased waste as it relates to the poor dimensionality of those avatars in relation to that rigged system. Now, how does this play into generative AI, right? If you don't have the right mass, if you don't have the right placement of everything. Just the same way as the fabric people had said in their due diligence process, you need to have lab controls. Where are the measurements? How high from the floor are the measurements? What's the circumference from that zone? You have to have aspects of standardization within your own company when you're measuring. And then the brand, um, it results in actually in generative AI as a way to control the spec, right? So if, if we get to a point where we can actually engineer in AI, specs are very important in that process. And then brand integrity overall does decline if we're not paying attention to all those variables and if we're relying upon AI that hasn't been tested. For mathematical integrity. So this is images from our current rig system. We don't give away the full system that's underneath it because it does do more than just this. But again, touch where your neck bends and see where that neck bends or see how your pelvic area tilts. Which this is kind of an important thing because it does allow us to deal with more postures than just one. So I could have a fit model in the digital space, but use the rig system to then position it through age. So if I'm working for like Chico's, I could go from 45 to 75 and figure something out for all those age group markets. And this is an example of that. So I took the, the rig system and I was able to apply it to an OBJ body and put it into positionality and get curvature of an aging body, which typically, again, related to the spine and the position of the, the 4 joints that we typically get in our programs, you cannot do that. You can't adjust your pelvic tilt, your buttocks, and even asymmetrical knock knee poses. So this is our process, our due diligence process. We do, for our standard, we use over 280 points of measure to match with the body, which is really important. We also work backwards. So in order to vet that we did the correct job, we actually started with a sloper and then moved the sloper into testing, and that seemed to work much better. I also worked with a chiropractor on the spinal placement, so we didn't want to put all of the vertebrae in because that's just too much data, but we put some in, and in that that space, we worked with a chiropractor to say what kind of shoes would impact certain joints in the spine, and then we use that to place the, the joints mathematically in relation to the body. So it's a fusion with biomed, if you will. And then we also configured shifts of gravity as we age. So as we age, somebody's center of gravity might be here versus here, or here. So you have to have 3 points of center of gravity that you understand where those exist in order to make it all work well. And this is us taking the sloper that we developed. Now we did not fit her at all, so this, she's standing there in something that we did without testing her. So there was no pre-fit. We just did it off of spec measurements and her data that she gave us. And then we built a form for that. And did it on a dress form as well. And then we took that sloper and moved it into CLO and tested it on the in-program avatar in CLO. And this is the ASTM, the C38 standard. Then we set up the morphology in Browzwear to have the same specs. And that's the instance that we got with the same pattern. And we made sure all specs were correct. And the last one, we all know that one. This one we couldn't into the, the poses that we wanted to do for testing, so that, that's a little bit out of the lab test, but it still shows obvious fit issues in relation to the sloper. And then we draped that sloper on the dress form in relationship to the fit model that had the same specs we requested. And then this was the outcome of that process. Now we rendered it directly in Cloak to show what we get on the technical end. So it's not in a marketing end. And if you notice her posture, it's not something that was straight up and down. She kind of had more of a forward tilt because we're dealing with the, the metrics of her body. And I've been working with Little Miss Tennis, which is the longest-running tennis company in the US for children's tennis, founded in 1973. And they're very much known for their quality, so their items resell sometimes for $10 less than what it retailed for. Uh, so they have a long life in their product, and accuracy is very paramount for the brand because we do dress all of the president's children for the last 35 years. That said, um, we started using digital avatars and testing with that process in 2018, and this is where we're at now, uh, and this would be a size 4-5 child. In which case, it fits and works better, much better than the Albinon one that we usually get. Because we can control its level of athleticness, right? And in this process, we're now at 50% less samples. So if it's a new style that we're really questioning, we will run it through a sample process. But overall, everything else is ideated and approved on in 3D, and then we move it into production. And that gives us about 8 weeks more of improved productivity over time. So for a moderate-sized brand, that's quite good, I would say. We have worked with bigger companies that we cannot name under NDA, but in that NDA process, we've been able to accurately approve Pant Fit that has outsold everything, and it was for a dress house. So it outperformed the dresses. And this is a little case study example of what it's like for a new brand. So if you get a new brand that wants to execute an idea they see on Instagram, right? Now Instagram model is just so appealing to them, and they get You give it to a freelance designer, and that freelance designer only has the tools that are available on a very limited budget. What do they do? They'll use the in-program avatar, right? Or they could use something like ours. And in that process, have stretch marks and flesh and tissue showing on that avatar. So when somebody gives us a design like this, we can say, OK, what size range are you going to sell it in? And for us, we design backwards. So we are designing on the plus or the largest size first and then scaling down. Because generally speaking, if it looks great on a plus size or a large size body, it's going to scale. But everything can look great on a size 2, but it doesn't always scale, obviously, to this picture. So when I had to create create the solution for this customer who has an outdoor motocross brand and deals with very petite women who ride motorbikes and who have muscle mass and generally babies. I came up with a bikini that was high-waisted with sheer insets, again working with the textile he wanted and providing the past body, because again, we put it on the plus-size— his avatar we could not show, but this is essentially what we arrived at. And he had over 220 images from Instagram that he said, please execute. And I said, let's, let's show you first how they look in scale. And out of that, he only executed a you know, up to 30 of them really. So amazing difference, right? And then this is really us doing an ideation to the muslin pattern. It on a size 8, which is the same as the sloper, and then on a size zero model shot for Ecom. So we were able to go up and down in the process and keep everything pretty realistic throughout in terms of mass and size. And we also comply with law and regulations, so we have worked with different standards groups, of course IEEE, and the 3DRC, the 3D body processing, which I'm a part of. Uh, we also work with the due diligence laws and Corporate Accountability Act, where we keep all information privatized, uh, and of course sustainability laws. So if you want to get in contact with us and talk more about it, we are still in the process of putting the rig system through CLO and Browzwear, which they have to do a little conditioning to take our, our stuff in. But ultimately, we hope to have that done by August of this year and have it ready for people to test as well. And we also do custom avatar work in the instance that you don't need the rig, but you need a customized avatar for size and mass. We go all the way up. And if you have questions, please feel free to ask. Any questions?

Thanks. Great presentation. That case study that you presented about the girls, tennis. When you got to that, to those final results, to have 50% less samples or more productivity, how long did it take? I mean, from beginning to end to get to those results?

Well, I honestly had to deal with the tools because I knew that they were wrong. Like, I literally have saved emails of me asking different 3D companies to change their avatar rig. And the question— the problem was, was that it was too great of a workload for their team to do. And then the other one was that we were doing it wrong. So in that time, it's taken a while, quite honestly, because the tool was wrong and I had to figure out how to how to rig and how to weight all— like, I literally was up late at night teaching myself how to do that stuff in order to create the solution, in which case it works very well. But the bigger feat now is its compatibility with computing in other programs. And that's where we have to have kind of standards for all those different things. So anything that's a foundational tool has to be accurate for us. And if we're asking it to go into the generative AI space, it must be accurate. Hi.

Great job on the rigging and adding the extra elements. I'm curious about male riggings and— size run avatars and specifically going up in sizes up to 60-inch chest. Have you been doing that work as well?

Mm-hmm. Yes. Okay. Yes, we can cover off all sizes, including down to children's. You saw the pelvic tilt part on the female. We do have something very similar to the male. Yes, moment.

I've expressed a couple of times here, talking to different people, how over the years I've observed changes in women's body posture. And I call it the Jane Fonda era versus the millennial era, where we were taught to walk around with our pelvic tilt back.

Yeah.

And now the pelvic tilts are forward with the butt out. And that makes a big difference on your demographic that you're building your avatars for. So just wanted to throw that out there.

Oh, yeah. No, that's why we put it in. But the men do also have a same feature. And then in terms of the genitalia, we do have a form we're working on right now that's just more for underwear, given that there's companies like Balls and Sacks that separate things out. And in order to effectively do that, you do need that anatomical correctness in order to create that cycle. But, you know, that would be considered nudity under the internet laws, so we don't show that. But yes, we're working on it.

That's it?

No more? Anything else?

No?

Okay. Perfect.