Generation of Graphic Design Models using Stable Diffusion

Generation of Graphic Design Models using Stable Diffusion

Deepak N R1, Shaik Jannat Al Firdaus2, Shayan Shaikh3, Syed Tajamul Shah4
1Professor, 2,3,4 Student,

Department of CSE, HKBKCE, Bangalore, India

ABSTRACT

The generation of AI Images in recent times has become more mainstream and this technique is quickly making its way into various industries that need asset creation and image generation. Although it is still somewhat limited, AI image generators can create breath-taking photorealistic images from complex prompts. Diffusion models can take in some user input, and some noise, and create content like what it has been trained on. For such cases, the diffusion models are trained using images, and input to the model. Upon passing a user prompt or input, we have the model produce a related image based on the input. Basically, a generative model using the diffusion method can generate images by adding Gaussian noise to the training data, randomizing it, and reversing the noise processing to restore the data. Diffusion models are parameterized Markov sequences, trained with different inferences to produce a picture that fits the data after a period.  Using this technology and resources, we infer to create our model that will make use of all these prompts.

Keywords: - Diffusion Modelling, Graphic Design, Stable Diffusion, Image generation, Machine learning.