Interactive Neural Painting

In the last few years, Neural Painting (NP) techniques became capable of producing extremely realistic artworks. This paper advances state art in this emerging research domain by proposing first approach for Interactive NP. Considering a setting where user looks at scene and tries to reproduce it on painting, our objective is develop computational framework assist user’s creativity suggesting next strokes paint, that can be possibly used complete artwork. To accomplish such task, we propose I-Paint, novel method based conditional transformer Variational AutoEncoder (VAE) architecture with two-stage decoder. evaluate proposed stimulate area, also introduce two datasets. Our experiments show provides good stroke suggestions compares favorably art.