Пошук академічних текстів

Інформація × Реєстраційний номер 2122U006296, Матеріали видань та локальних репозитаріїв Категорія Препринт Назва роботи Mapping Materials to 3D Texture Field using GANs Автор Novosad MarkiianNovosad Markiian Дата публікації 01-01-2022 Постачальник інформації Український католицький університет Першоджерело https://hdl.handle.net/20.500.14570/4436 Видання Опис 3D Texture Synthesis is a broad research field which lets content creators get visu- ally satisfying look of the 3D object with minimal effort. However it is not so easy to achieve satisfying result coupled with computational efficiency. Many of the pro- posed methods are either computationally expensive and inefficient, or not capable of generating realistic visual appearance. To address this issue, we propose a new method – Mapping 2D Materials to 3D Texture Field using GANs. This method is based on Neural Implicit Representation network, thus able to internally represent a Texture Field without a need for storing additional information. This ability lets our method to be computationally inexpensive, theoretically being able to render the texture in real time. On the other hand, due to Generative Adversarial training strategy, our method is able to achieve highly realistic, visually satisfying looks. In this study we will describe our approach in detail. Додано в НРАТ 2025-11-05 Закрити