Creative ML
Fall 2023
To integrate Machine Learning into the fashion design process, collaborating with AI to function as both a tool and a wellspring of inspiration for the creative journey.
Can Machine Learning Replicate Fashion Design Process?
-
Let AI imitate a fashion portfolio based on an existing collection by Stable Diffusion and training Gan model
Creating a fashion collection is a complex and time-consuming endeavor. Every garment embarks on a meticulous journey, evolving from initial sketches to the ultimate creation through numerous iterations, adjustments, and meticulous fine-tuning.
Artificial Intelligence (AI) can swiftly extract essential elements and produce a myriad of images. Therefore, I am considering utilizing AI to create a fashion portfolio based on fashion collection works I have undertaken over the past three years.
Here are training results obtained through Stable Diffusion, a tool that can broaden designers' perspectives and enhance the ideation process.
By incorporating key words and images, it significantly improves designers' efficiency and opens up possibilities for more design alternatives.
Researches and Develop Processes
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
This process can rapidly produce images with a highly polished finish, resulting in some outputs that are both unrealistic and amusing. Machine learning has the ability to capture a similar vibe, yielding outcomes resembling creative collages. The unpredictable results can serve as engaging advertising images for the design process.
Some designers feedback:
-
Cool that you're trying to incorporate your own work into the machine: Very apt way to exploring the relationship between human labor and AI! Super interesting to see the different humans generated from the prompts! and i think it would be cool to explore if AI/ ML can replace each individual portion of the process as well. eg making patterns
-
Like your project, really solid and useful. I wish it could evolve into a website or an open-source platform, enabling designers to upload their works and help them with their garment design process.
.png)
Input 572 images from Stable Diffusion to train my own Gan dataset. by using Google Collab.
.png)
.png)
The abstractness of the Gan training process could be a great tool for the designers to develop shape and silhouette.
Materialize in reality: Print out the image and utilize tracing paper to enable fashion designers to sketch silhouettes and garment designs inspired by the GAN training image.
This process aids designers in creatively developing silhouettes, accelerating the workflow, and enhancing overall productivity in the design process.
- THE END -
Employed Python algorithms to generate text.
.png)
.png)
.png)
.png)
.png)
.png)