Machine Learning in Design

The day we all once thought was fiction has arrived. The age we spoke as if it were a million miles away. It’s definitely a boon for all of us. I presented a small talk on Machine Learning in Design.(cause everyone is talking about it.) I cover majorly technology that is around designed.

Yellow Slice - Machine Learning in Design
is it true?

Personally I think machines and their intelligence can coexist together to help us reach peaks like never before. Before I start I will really quickly run you through what they call the basis of machine/deep learning.

Yellow Slice contact

Imagine Hard Disk is the brain and the CPU is the processing unit. Now as we know it the brain are made up of many many neurons which are triggered each time we access our brain, using the same model they have created something called neural networks
Source:
https://en.wikipedia.org/wiki/Artificial_neural_network

These neural networks help us to recreate various aspects of our brain and so there are many kinds of neural networks.
1. Feed Forward Neural Networks (LSTM)
2. Recurrent Neural Networks (RNN)
3. Generative Adversarial Networks (GAN)

Right now I will focus only on GANs (Generative Adversarial Networks).These have been recently applied and works brilliantly with images. There have been many use cases for the same and here are a few of them.

  1. Object Finder — After learning several images of cats the bot can easily detect a cat in an image.
    Source: https://github.com/llSourcell/tensorflow_image_classifier
Yellow Slice - Machine Learning of Object Finder
Oh! Hello Cat.

2. Text to Image — This is really cool stuff cause now after training the machine on images with captions it was then able to generate images directly from the text. In the below example you can see after learning images of flowers and birds the machine in 2 stages could generate images exactly as per the inputed text.
https://github.com/llSourcell/how_to_convert_text_to_images

Yellow Slice - Machine Learning of Text to Image
Machine : Yeah I made it!

3. Image Manipulation — I love this the most cause it helps designers as it directly helps image rendition and morphing piece of cake. So based on the drawings the machine can directly understand and create an image for it. The source is still unknown but it’s for real. Below is a screen grab from the actual code.

Yellow Slice - Machine Learning of Image Manipulation

Do like and Share 🙂

  • Sunny Padiyar

FAQs For Machine Learning

What can Machine Learning do for Design?

Machine learning can be used to automate the design process, making it faster and easier for designers to create new designs. Additionally, machine learning can be used to create designs that are more personalized and tailored to the needs of individual users. Finally, machine learning can be used to improve the quality of designs by identifying errors and flaws in designs more quickly and accurately.

What do Generative Adversarial Networks do?

In a GAN, there are two neural networks: a generator and a discriminator. The generator creates data that is similar to the training data, while the discriminator tries to distinguish between the real training data and the generated data. The two networks compete with each other, and as they do, the generator gets better and better at creating data that looks real.

Why do we need GANs?

Generative Adversarial Networks, or GANs, are a type of artificial intelligence algorithm that are used to generate new data. They can be used to create new data for training machine learning models, for example. They can also be used to generate realistic images for computer vision applications or to create new audio for speech recognition applications.