DragGAN AI
Introduction
AI is one of the most spoken tech topics around the world now. With AI like chatgpt our lives have become much easier will continue to do so. (let's just ignore the fact that they be stealing out job-)
AI has once again helped us, this time in the field of photo editing, which could no longer be as time taking with newly released DragGAN AI.
What is GAN?
(TL;DR: GAN generates near to perfect fake images)
Before going into what is DragGAN, let us first understand what is DragGAN.
Generative Adversarial Networks, or GANs for short, are an approach to generative modeling using deep learning methods. Let's assume a machine, which has 2 parts, generator and discriminator. We feed discriminator with some images say real cat images, we feed it until it is able to discriminate cat with some other image say dog. Now that our discriminator knows how cat looks like, we start our generator which generates fake cat images. If our discriminator is able to identify the fake, our generator upgrades itself and generates better fake, if our discriminator couldnt, it upgrades itself into identifying cats lot better than before. This process iterates a lot of times until our generator is able to fool discriminator completely. This exactly happens in GAN. Hence this is used in DragGAN to "edit" the images.
What is DragGAN AI?
DragGAN stands for “Drag-based Generative Adversarial Network.” It is a type of AI model that can generate realistic images based on your input. Unlike traditional image editing tools that only distort or crop existing pixels, DragGAN can create new content that matches your intention. With DragGAN AI Tool, you can regenerate, rotate, and transform elements within your photos, giving you complete control over the editing process.
Due to such flexible and efficient features, it is highly hyped up to the point it is now called as "photoshop killer".
Intro on how to use
It's very simple to use.
How does it work?
It creates entirely new content that fits seamlessly with rest of the image using GAN. Let's say you manipulate someone's head to rotate it DragGAN will create unseen facial features from original image like the ears or teeth making it all appear incredibly realistic, it also adjusts the image's lighting and shading to enhance it's natural look. Best thing about DragGAN is it doesnt have any restriction, it can manipulate any sort of image. Also the impressive speed and efficiency stand out .
What was the purpose behind creating it?
"Synthesizing visual content that meets users' needs often requires flexible and precise controllability of the pose, shape, expression, and layout of the generated objects. Existing approaches gain controllability of generative adversarial networks (GANs) via manually annotated training data or a prior 3D model, which often lack flexibility, precision, and generality. In this work, we study a powerful yet much less explored way of controlling GANs, that is, to "drag" any points of the image to precisely reach target points in a user-interactive manner, as shown in Fig.1. To achieve this, we propose DragGAN" - taken from official paper in arxiv
In summary, the researchers present DragGAN as a powerful and less-explored method for controlling GANs, offering users the ability to interactively manipulate generated images by dragging points to achieve desired outcomes with flexibility and precision. The paragraph does not exactly mean "Photo editor" rather an image manipulator to control various aspects of the image such as pose, perspective, shape etc..
Is it actually "Photoshop Killer"?
From what I have seen, I would say it is not really a photoshop killer. DragGAN was released recently and is still in developement process, as of now it cant edit images as good as photoshop since it's still in its early days. But with years to come it's definitely going to improve a lot, but wont completely replace photoshop. (Maybe there will be another AI made specifically to replace photoshop?)
Limitations:
Since it's a machine learning model, it isnt perfect as of yet.
It needs a wide range of data to make near perfect editing (as explained in GAN section, more data feed = more realistic image generation)
Say you only feed data of laptops, it cannot edit images of say a mouse.
If it doesnt have this wide range of dataset, it could find problems in generating some images causing visual errors.
(I hope they fix this)
Anyways-
Conclusion:
Despite these limitations, it clearly is highly advantageous when it comes to photo editing.
The more users/images/data it gathers, the more near perfect editing it could do.
Draggan is the result of years of research and developement in machine learning, graphich, computer vision, human computer interaction and many more.
Sources:
maarat_ai on youtube
IBM technology for GAN
Comments
Post a Comment