Portal de Periódicos da CAPES

Creative AI Using DeepDream

2023; Springer International Publishing; Linguagem: Inglês

10.1007/978-981-99-7093-3_19

2367-3370

Rakhi Bhardwaj, Tanmay Kadam, Shubham Waghule, Sahil Shendurkar, Bhakti Sarag,

Creativity in Education and Neuroscience

The term “creative AI” refers to a branch of computer science that automates the creation of unique pieces of art, music, and literature. It allows technology to replicate and help human beings in the creation of unique and imaginative items. DeepDream, a computer vision tool, uses a deep neural network to find and improve patterns in images. The approach works by submitting an image to a trained neural network, which recognizes specific elements or patterns in the image and amplifies and enhances them. The current DeepDream algorithm’s usage of a single image has the drawback of potentially producing overfitting, limiting the algorithm’s ability to produce images that differ from the input image. As a result, the variety and originality of the developed graphics may be constrained. To get around this problem, DeepDream algorithm was trained to use multiple images or a combination of images and noise, resulting in more varied and inventive outputs. This allows us to create imaginative films as well as numerous visuals using just two images.