【Will the AI refuse to accept pattern data generated in 2D CAD drawing software for further processing?】

Generative AI is a hot topic of the moment, and we all know that everything AI can do requires a process of learning from the data we provided.

Let’s start to think about AI application in clothing pattern generation, and we will realize a very serious and logically important issue.

The ultimate forms of both virtual clothing and physical clothing are 3D. And the common source is clothing pattern data.
If the pattern data provided for AI learning is generated from 2D CAD drawing software, then the source data is not generated in 3D state, it is not logical that AIGC can play a role in this situation.
All kinds of existing picture generation, can only be skin, and can not achieve the real meaning of DX❗️

2D and 3D are completely different dimensions, so no matter how AI learns, it can not make perfect transformation.
AI need to learn that a combination of 2D patterns can be generated with different shapes based on cutting lines on the same 3D garment model.

If there were no massive pattern data generated like our method to provide for AI learning, there would never have been an AI product that could directly restore and convert a picture of a garment worn on people into pattern sets that could be produced automatically or semi-automatically❗

https://lnkd.in/gprpAubA
https://lnkd.in/gvggGZQe
https://lnkd.in/gTufNWJ9

According to the content of an interview with Professor Geoffrey Hinton at MIT Technology Review on May 3, soon AI will be able to conduct thought experiments.
Will the AI refuse to accept pattern data generated in 2D CAD drawing software for further processing❓🧐🧐

21st May,2023
#design #digital #digitalfashion #digitaltwin #3dmodeling #sustainability #apparelindustry #metaverse #fashiondesign #AIinfashion #AIGC #business

A Digital Fashion Show Workshop with Sichuan Fine Arts Institute

This is a digital fashion show completed by Sichuan Fine Arts Institute and SdibiT., Ltd. Fashion students have their own design ideas. But how to turn that idea into a feasible design and make it true? Here SdibiT plays its role. We make it come true and make it more vivid in VR display.

Digital Fashion/Digital-twin Fashion/DigitalTwin-Matching Fashion

Digital Fashion/Digital-twin Fashion/DigitalTwin-Matching Fashion, are three different concepts.
Today, there are overseas brands coming to consult: What are the highlights of your #metaversefashion video on Fashion Week❓
https://lnkd.in/gErwiJ_M

Our video is dedicated to achieving DigitalTwin-Matching virtual display.
It means the physical clothes that are truly delivered to consumers are not only the appearance, but also the wearing effect matches the virtual display as much as possible❗

 
There is still a big difference between the costumes of animation or game characters in pure visual entertainment:
1. Garment modeling
The patterns generated traditionally from 2D Cad and provided by the brand, are not directly used❗
We verified patterns by our data processing system with 3D calculation optimization and reconstruction.
https://lnkd.in/gy3zw2EH



2. Vmodel modeling
The characters’ body shape is also the output of our system processing, not a detached supermodel or anime game style body shape❗
https://lnkd.in/gEaPRjcb


Digital Fashion ❓
Digital-twin Fashion ❓
DigitalTwin-Matching Fashion ❓
Just tell SdibiT your application scenario and purpose, then you will get a one-stop arrangement.


March 29, 2023

【How the ordinary ready-to-wear business balance high level design and pattern making cost with revenue? 】

生地の色柄が毎年違うだけ..と不満を感じたことのある人は是非お読みください

This article is specially for people who are not satisfied with changes limited in fabric colors and patterns.

The way to generate 2D patterns that addresses Fit&Size Issues is very different from the traditional way of making clothes today.

First, create a 3D prototype with a silhouette that ensures comfort.

Second, create 3D Garment Models according to different requirements from customers or different sizes. Then generate 2D patterns.

Using our company’s method, it takes more than 3 times working hour to make 2D patterns corresponding to #DigitalTwinMatching.
That is to say, it is much more difficult to make patterns achieving good Fit&Size.

At the end of last year, we accomplished a job of making garment 2D patterns which realizes DigitalTwin-Matching according to designated design drawings. This is the most difficult job so far and it charges more than $12,000 for 1 Look& 1 Size considering time and labor cost.

Brands that can bear such high pattern making cost are limited in #luxury brands like Louis Vuitton, CHANEL, Valentino, Versace, etc.

Then how the ordinary ready-to-wear business balance such high design and pattern making cost with revenue?
Learn from other manufacturing industries.
1. Maximize the use frequency of universal design:
Reuse 3D designs or 2D patterns that have already been developed, and increase production volume by using a 3D design developed once for multiple items.


2. Minimize design changes:
Minimize the parts subject to change to reduce additional costs associated with design changes.
In other words, standardize the design.
 
Secondly, in the fashion design section, turn the priority work from making 2D patterns to management of component design list.
At present, the traditional working mode is that the design section starts from scratch to operate according to different specifications every time. Data information is isolated and fragmented, resulting in a great waste of repetitive time and labor.
Working mode of the design section needs to shift from “centered on design /making 2D patterns” to “centered on BOM(Bill of Materials) management “.

It’s not the standardization of work.
It’s about regularizing the design.
Promote modularization while meeting the specific requirements of different customers.

Matching and combining modular design parts is our Mix&Match-Technology solution.

Here is a cost comparison list.
If the current design cost of 1 look is 100.
Adopting our solution, the cost of 1 look would be 300.
But if you reuse the design with modular management for 10 times and make 10% design fine adjustment every time, then the cost would be 300+ 30×10 =600, and 600/11 is less than 60.

Modular design of best-selling products can greatly reduce the cost and help ordinary ready-to-wear brands achieve high design level as top brands.

8th, March

【Some parts of technology logics of AI search tools are similar or even same. Can any giant company plus α to differentiate itself? Starting from AI fashion with us is promising】

After the company OpenAI, funded by Microsoft, released #ChatGPT and combined it to search engine Bing, #Google also announced its plan to integrate Bard, its self-developed AI search tool similar to ChatGPT.

For users, since platform tools have similar functions, will the platform that can provide in-depth services on this basis be more attractive? The demand of fashion design cannot be missed!

AI technologies have applied in fashion industry in many aspects. For example, natural language processing has been used in conversational commerce, computer vision has been used in smart mirrors, and to be noticed, generative model as fashion designers. Google, #Amazon, and the Chinses company #DeepBlue all released their AI fashion design application around the year 2016 and 2017.

However, why these AI fashion design tools did not been put into use widely?
Having a lot of AI generated design pictures in hand but have no efficient way to create pattern data to turn them into actual production is useless!
Most users want to actually wear them, not just enjoy the pictures, don’t you agree?

As mentioned in the previous submissions, the point to notice is that “Generative AI” learns from image data that already exists.
Our company SdibiT has millions of pattern pieces which can be matched and combined into a new style. We call it Mix&Match technology and it is a solution to AI generative fashion.
(https://lnkd.in/gywZcCh8)


However, to maximize our data and technology application, we’d like to work with a professional AI team. We are going to figure out the algorithm of automated pattern matches together and generate as many styles as possible which all have the adequate data for production.

Is it an exciting work? Any team has the ambition to shake the fashion industry is welcomed to contact us🤝🤝

4th.Mar,2023

【Our solution is not the competing version of #CLO, #Browzwear, #Optitex, #Style3D…Our relationship is mutual supplement】

As mentioned in the last submission, the consistency of data is fundamental for DX in any industry, and the fashion industry will be no exception❗
Should these enterprises integrate our scheme into the system for comprehensive application❓
Since it will truly innovate and benefit the design department of fashion industry and provide the industry with the true DX process❗

In China, our company is invited by the design school of university in the video below to hold workshops regularly. Let the teachers and students understand the basic digital design thinking method and process of design data processing by creating their own digital garments together.

The purpose of this workshop in the video is to experience the whole DX process. Therefore, left behind our company’s requirement of DigitalTwin-Matching temporarily, teachers and students can first enjoy the creation freely.

Note the first twenty seconds of the video. The method is obviously wrong if considering the actual wearer.
The rigor of phygital demands you not to draw the curve arbitrarily. How to think and deal with it, is the teaching content of our next workshop.

However, the great players of the future have realized the significance of designing directly on the “3D Digital DRAPING” according to their own idea and obtaining digital pattens which can restore the design in real time.
One small step today, one big leap tomorrow. The future is in sight👍

27th.Feb,2023

【3D Digital Garment Model should be the starting point of the design and pattern making process, rather than just the output result of the process】

Clients who get to understand our solution for the first time often have the following response. “Ah, comparing #CLO#Browzwear#Optitex#Style3D, etc, your DX solution is in the reverse order.” And I would say every time that we did not deliberately reverse, our solution just correct the process to the right track❗️

Let’s think about it together. The clothing making process 150 years ago, is just like the modern haute couture customization process. Each piece is made based on the customer’s body and confirm the right fit&size, then the 3D prototype obtained on the real body is converted into drawings of 2D pattern, being sewed finally.
In the era of industrial high-speed mass production, the process gradually deviated from the original route due to time, efficiency and cost, and the most important reason, there was no 3D digital design tools.

Strictly speaking, from the perspective of other industries such as architecture, shipbuilding, etc., #design drawings and 2D processing drawings are different things❗️

2D pattern maker’s job is not creating design drawings but generating CAM data for execution of processing output.
The popularity of 2D pattern making software in the era of mass production is because it enabled efficient management of the CAM data used in production line processing.
With the development of #digital technology, high efficiency, precision and low cost can be achieved simultaneously while the steps are corrected back into the logically right order.

Let’s think further more together. If one customer likes the design draft generated by AI and decides to buy, the seller just need to make the physical clothing faithfully restore the design while ensuring the right fit&size of the customer.
The core advantage of our solution is the ability to quickly convert designs into production pattern and achieving DigitalTwin-Matching at the same time. After this, the job can be done easily. Just sew the specified fabric according to the pattern.

So here comes the problem. At present, the pattern generated by 2Dcad is first used for 3D modeling and then be sent to the customer for confirmation. If we receive modification requirement, we have to get back to 2D pattern to do modifications. Except for some simple basic styles, will this kind of workflow definitely put the 2D pattern makers into hysteria❓
The workload do not decrease but increase because they have to use the modeling and rendering software to make 3D views of patterns they made and send to customers for confirmation repeatedly. What’s more, this kind of display only shows the appearance of the garment, it does not show the Fit&Size state at the same time…
(to be continued…)

27th.Feb,2023

【In the next 3~5 years, the value of the production pattern or nested grading pattern which started from 2DCAD will decline rapidly.】

  

The special TV series about underwater archaeology broadcasted yesterday is very interesting. By introducing the latest 3D simulation and modeling tools, archaeologists can make the most effective excavation plan and make records with millimetre accuracy while easing their physical workload. In this way, they preserve human historical heritage.

One doctor of the archaeologists emphasizes, the excavation work is also a kind of destructive activity in a sense. The original feature of the heritage maintained during the long period of more than 2000 years must be accurately and faithfully recorded in the state of 3D models. The heritage of mankind should also be in form of digital asset that can be faithfully CG-restored and handed over to descendant.

The stones of those hills may be used to polish gems.
If there is no millimeter-accurate 3D model to record the data in advance, how can we accurately restore the original spatial position, angle and distance of the archaeological site by memory and feeling once it has been touched?
In the same way, if the editing and modification of clothing pattern has no identification of the design position and shape in 3D state in advance, and no corresponding operation, how can we achieve the faithful 3D restoration effect by directly making the pattern in 2DCad software by virtue of memory and feeling?
 
While it has become common sense for other industries to use 3D models as the starting point for process, the fashion industry must keep up with the times.
 The change of concept is the first priority!

The benchmark for the value evaluation of pattern data assets in the future will become whether they come from digital 3D garment models instead of 2D CAD software.

24th Feb,2023

【Fashion designers who only know how to draw a graphic design will no longer be needed】

“If AI generates a large number of fashion design drawings just like the hot topic chatGPT does, what will SdibiT, ltd do to use these drawings?” A customer who visited us today asked a question.

This question reminds me of what Michio Tomii, CTO of our company, has been saying for nearly ten years, “In the fashion industry, designers who only know how to draw a graphic design will no longer be needed.”
However, his inference ten years ago cannot be proved obviously and get understanding and recognition of people in the industry because the tool which can turn the intangible data into tangible and visual information has not emerged.

Since last year, tools like “AI Art Generator” (txt to Image: DALL-E, Midjourney, Stable Diffusion… etc.) are starting to emerge, making it easy for anyone to generate stylish design images. And this kind of technology will evolve in the future.

Currently, AI can generate Virtual Show-On, but cannot output 2D patterns that can be reconstructed to the 3D-image.
To achieve DigitalTwin-Matching, highly accurate pattern data in accordance with Fit&Size is necessary.
There is no denying that AI generates design&pattern at the same time may be possible in the near future.
The point to notice is that “Generative AI” learns from image data that already exists.
 
Our company SdibiT’s approach is illustrated as the chart.
Starting from module design, we call this Mix&Match→Combination→Compilation.
By editing small pieces of patterns generated by multiple designs as parts, large amounts of pattern data can be generated.
In fact, as a complete form Phygital after integration, only a very small number of it is made public, most of it is kept internally.
That is to say, we have a considerable amount of data assets in our company, which are raw data that cannot be utilized by other companies through Generative AI.
Among the massive design images generated by AI, there may be actual clothing that is not acceptable to consumers or the actual situation cannot be completed exactly the same as the picture, the next candidate design will be generated by the “average consumer”. Based on this, multiple different candidate looks are generated while creating a design data that achieves DigitalTwin-Matching.
The “average consumer” can choose among these candidates.
This process leads to true humanity and individualization.

“In the near future, everyone will not only become a fashion designer, but also be able to confirm the effect in 3D. Fashion designers who only know how to do 2D painting will almost disappear sooner or later.” Mr. Tomii always said.

21st Feb,2023