AGL’s bright future and the first attempt to use AI in fashion moving imagery
AGL is one of the liveliest and brightest creative brands for female shoes in Milano. New York Style Guide has been following up, close and personal the evolution of this brand, under the lead of the three creative, imaginative and talented sisters Sara, Vera and Marianna. And every season they manage to do something out of the box that is absolutely worth the mention.
In a world well dominated by men designers, it adds to the value of the product, that it is designed by women, for women. The level of comfort when we talk AGL is hard to compete with. But the thing about AGL, is that they have a sparkling creativity and they literally use Fashion Week as a hub to promote something new, not just design and portable shoes, but also for opening new visions in fashion business strategies.
Though they are not one of those brands that like the scream and shout philosophy, they are incredibly innovative. They act as focused professionals in an environment where, at times, screaming marketing, is a cover for a very thin content. The AGL brand shows interest in the whole reality of fashion, and wants to investigate opportunities in marketing in a very Milanese way: using every occasion as a good opportunity to become something new and something better.
For that reason, there is always a bit more to talk about, than just plain design, when it comes to AGL. There’s philosophy, there’s technology, there’s more than just shoes and when we look at marketing strategies integrated with new design, it’s always worth to consider what Sara, Vera and Marianna are doing.
Unveiling what the title already suggests, as first in the fashion business, the AGL sisters decided to introduce an innovative virtual project using artificial intelligence. Created by videographer Alberto Maria Colombo, AGL’s project uses AI through the GAN algorithm.
What is the GAN algorithm and what’s the use in fashion?
Trying to make everything as simple as possible, and may all experts forgive me for any over-simplification that might sound incorrect to them, GAN stands for Generative Adversarial Networks, and GAN algorithm is a class of machine learning frameworks and an approach to generative modeling. It is an unsupervised learning task that involves automatically discovering and learning the characteristics of an input data in order to generate a new output data that is absolutely credible. Bear with me.
There are two sub-parts that work together: a generator and a discriminator, let’s see if I can create an example connected to the fashion reality. Let’s say that we want to use GAN to create the image of different shoes.
This is what the generator does.
Let’s take a shoe as input data and let’s call it S. S comes with a lot of attributes, like it’s a blue suede flat shoe. These all translate into numbers, say 01 for blue, o2 for suede 03 for flat that we collect under a different letter, a small z, that can be written z (01, 02, 03).
Simplifying: S(z) where z(01, 02, 03) means blue suede flat shoe.
Hence z is all these attributes, and the numbers correspond to the specifics. We want the Ai to generate something else, like a completely different shoe. Unsupervised learning means we don’t teach it what and how to learn, we let it decide without our control. So through the generator it’s learnt that S is a shoe, and that all numbers in z mean a different specific, 01 means blue, 02, means suede, and 03 means flat. At that point the Ai in deep learning can give values to a new shoe, changing the combination eg in an arbitrary 04 meaning red, 05 meaning patent leather and 06 meaning stiletto.
Simplifying: S(z) where z(04, 05, 06) means red patent leather stiletto shoes
So the generator uses a code to generate different images of shoes.
At this point we need to look at what the discriminator does.
To define it, I’d say that the discriminator is a guide for the generator, id est it guides the generator on what images to create. It acts to understand the differences between real and generated images. The discriminator’s task is to answer the question “what are the features that make an image real?” and send the feedback to the generator.
To get there, we give the discriminator a training sample (an image, a photo) to study, and he uses the informations in it to compare with the image that comes from the generator. To use the example, the generator knows that s(z) where z(01, 02, 03) means a blue suede flat shoe. But how does it create it to make it look real? The discriminator has an actual picture of the shoe, which is the training sample. The generator sends its version to the discriminator, which in turn compares it to the real picture. If it spots the differences and can say that what the generator did is a fake, it sends it back, and then the generator continues trying, until the discriminator can’t see any difference between the generator’s work and the training sample, id est, the original picture.
So the discriminator has an important role, when the generator creates its new images, the discriminator is like the demanding teacher that sends the homework back to the student, until it’s perfect and the student tries to solve the homework finding his own way (unsupervised learning).
Hopefully I found -my way- that is also correct as possible to explain how GAN algorithm delivers generative models that create realistic examples.
Imagine the reality we live now, understanding the very basic of what this does, you can easily guess the potential in the fashion business.
When given an image-to-image translation’s task, think summer to winter scenarios, GANs generate photorealistic photos of objects, scenes, and people that we humans cannot tell that they are actually fake. The possibilities in fashion marketing (and not only) are potentially endless with different costs compared to traditional visual communication. AGL is pioneering with fashion moving imagery, and which better times than this to investigate Ai: in their video, human shape and shoes are recognized and mixed by an electronic machine to create an unpredictable flow of images.
Created also with an artistic allure, its meaning extends on how we actually create our own memory when we look at something new and how we do remember fashionable products.
Let’s finally see the AGL video:
Don’t see this as a competitor to the fashion workers. Ai is supposed to work together, not as a replacement of humans.
This is the team behind the creation of this video:
Photography and art: Alberto Maria Colombo
AI: Ümüt Yildiz
Hair: Marco Braca
Makeup: Alessia Bonotto
Models: Adja Kaba, Liva Reira, Jeske van der Pal
Assistant: Lorenzo Sampaolesi
Production: Annapia Lorenzi, Arianna Grava
Styling: Gili Biegun x AGL Shoes
Special thanks to Birgitte Herskind
(With superwoman: Ursula Giustina C )
AGL ss2021 highlights:
For spring summer 2021, AGL is talking to the new generations of women. In line with the positive spirit of Milan and the whole Fashion Week, AGL sees the light at the end of the tunnel: as a fil rouge all through the presentations, designers are vocalising their faith in the future. And possibly psychologically a “bright future” theme is a necessary memento not just for fashion, but for all businesses that have to cross a long and troubled winter.
Typical and specific of this brand, shoes are designed as said by women and for women: so the new collection is a concert of ideas shared considering the expectations of the new generations that translates into easy and versatile silhouettes in one direction, sculptural shapes crafted with clean-cut, or daring cut-off, soft weaves and leather nets in the other direction. That is because the customers are always represented by two major waves: comfort lovers and identity seekers. One answer in both direction, if the accent has to go on comfort or on extravagance, there is a shoe for every AGL customer.
AGL has a very defined and customer friendly strategy for shoe design, and the fact that they communicate with the clients, put them in a very interesting niche.
AGL motto is sober elegance and their care for quality translates into timeless contemporary pieces. We are talking about Italian artisans, so Made in Italy as sophisticated craftsmanship, use of high-quality materials and attention to details. Made in Italy is for clients that are interested in the whole idea of fashion. As many similar brands, AGL is for sophisticated women who have a high awareness and interest on how their shoes are made, where and with which level of care.
For their specific nature, apologies for repeating, but to women and made by women, the AGL design focuses on a sensitive balance between practicality and femininity.
AGL final customer can be summed up in a woman who:
- is cosmopolite
- cares about aesthetic and comfort
- looks for Made in Italy perfection
- needs comfortable day-to-night confidence
The three sisters made a small video, a talk on their new collection, that we can share with our readers. Who better than them to describe the new options?
Chosen by celebrities like Mila Kunis, Elisabeth Olsen or Dakota Fanning, AGL are known for being among the most comfortable shoes on the market: the value of these products is in their nature that is in line with the requirements of the Made in Italy: care for every step of its production, extra-value, product with a story, traceability, eco friendly, comfortable and practical, but aesthetically pleasant, varied to meet different expectations, in connection with customers, and made to last.
The spring summer collection is in line with the spirit of Milano Fashion Week: it’s bright and positive, as it’s in the very nature of hardworking Milanese people (born in the city or moved by call) to always believe in better, and it’s the reason why the city survived a lot of difficulties in its past, becoming a leading guide for the whole Italy, with higher success than other not only Italian cities. Milan has a strong sense of self-made dignity, self-esteem, devotion to work, efficiency and a form of immunity toward victimisation and pessimism.