Ullah Noor, Zabeeh  <br/><br/>
ENHANCED NOVEL VIEW SYTHESIS VIA DEEP LEARNING-BASED 3D GAUSSIAN SPLATTING / 
Zabeeh Ullah Noor 
 - 83p. ;  Soft Copy,   30cm. 
<br/><br/> 3D Gaussian Splatting (3DGS) has emerged as a breakthrough in explicit radiance<br/>fields and computer graphics which has enabled precise scene representation, real<br/>time rendering, and efficient novel view synthesis. This paper explores the evolution<br/>of 3D rendering and recent advancements in 3DGS, with a particular focus on<br/>different techniques for synthesizing novel views with the incorporation of deep<br/>learning architectures especially transformers. To enhance scene quality, this<br/>research investigates the integration of monocular depth information during<br/>rendering and refines the loss function to improve reconstruction accuracy. By<br/>incorporating depth information our method enhances geometric details by<br/>capturing intricate details and reduction of artifacts. The findings contribute to the<br/>reconstruction of 3D scene with high fidelity, offering insights to optimize Gaussian<br/>Splatting technique for more efficient and realistic 3D rendering applications. 
<br/><br/><label>Subjects--Topical Terms: </label><br/>MS Robotics and Intelligent Machine Engineering   
<br/><br/><label>Dewey Class. No.: </label>629.8