INNOVATION ON DEMAND: DESIGNING CUSTOMIZED CLOTHING WITH GAN

INNOVATION ON DEMAND: DESIGNING CUSTOMIZED CLOTHING WITH GAN

Prof. Neha Ghawate, Aniket Kale, Sharanu Belaki, Shailesh Salve, Vibhushan Pol

1Prof. Neha Ghawate I.T PGMCOE

2Aniket kale I.T PGMCOE

3Sharanu Belaki I.T PGMCOE

4Shailesh Salve I.T PGMCOE

5Vibhushan Pol I.T PGMCOE

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract – It is an innovative approach that aims to enhance the fissionability of a given outfit by proposing minimal adjustments that have a maximal impact. The approach utilizes a deep image generation neural network that is trained to synthesize clothing based on learned per-garment encodings. The encodings are factorized according to shape and texture, enabling direct edits for both fit/presentation and color/patterns/material. This approach also employs a web-photo bootstrapping technique to automatically train a fissionability model and an activation maximization-style approach to transform the input image into a more fashionable version. The proposed edits range from swapping in a new garment to tweaking color, fit, and the way an outfit is worn. These adjustments can include rolling up sleeves, making pants baggier, or adding accessories. The model suggests these edits by using automated metrics and human opinion to determine their success. Experiments demonstrate that Fashion++ successfully provides edits that are deemed fashionable by both automated metrics and human opinion. This approach presents an intriguing new vision challenge and an effective method for enhancing the fissionability of an outfit through minimal adjustments. With its ability to suggest minimal yet impactful changes, Fashion++ has the potential to revolutionize the way people approach fashion and styling. The project aims to generate high quality designs without having knowledge of technical and artistic skills in drawing. In recent years , GAN technology has been used in the fashion domain .We focus on a new problem of using artificial intelligence to generate new images. Basically, the dataset having more than 9000 cloth images of various designs is used. The main objective of the project is: 1)To generate unique and new designs. 2)To provide variety and effective fashion styles. The user can be introduced to a variety of choices of styles and design within a less amount of time. The manual efforts can be reduced . The merging of different style looks can generate a unique style. To make the system more effective, GAN technology and its modified versions can be used. It will not only give us the variety of results but also in an automated way within a required amount of time. The user interface will make the interaction feasible and attractive. The main idea behind the project is to merge two input images of two different kinds and generate their results by varying some parameters of them. The color contrast, sleeve varieties, style fusions, seasonal - traditional- occasional fashions from bottom to top can be achieved.

Keywords: fissionability, Design, Styles, Unique generation.