Researcher at TCS Research | MTech CSE'23  IIIT DELHI | AIR 923 in GATE(2021) CS&IT

github

Hanzalah Firdausi  • Peerlist

The Neural Style Transfer project is a web application that uses a Generative Adversarial Network (GAN) to apply the style of a chosen artwork to a user-provided image. It was developed using Flask, CSS, JavaScript, HTML, Python, and PyTorch and includes features such as user login, signup, and the ability to share the generated images on social media platforms. It allow users to choose from four different styles: Ukiyoe, Monet, Cezanne, and Vangogh. The project was completed by Hanzalah Firdausi, Lasani Hussain, and Ankit Kuswah as part of their final project for Data Processing and Management coursework (ECE 560) at IIIT Delhi. We thank Dr. Pravesh Biyani and course TAs for their supervision.

Neural Style Transfer

The project is done as part of and dataset provided by https://constraint-lcs2.github.io/. We aim to project a particular entity as either hero, villain, or victim based on the context it is mentioned in. We create and build multiple models for the share task analysis. The models are based on extracting the visual and textual context of the entity, aligning them with the entity itself, and using the shared feature vector for labeling.