Select a calendar:
Filter December Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for December
-
PhD Thesis Defense - Yuzhong Huang
Tue, Dec 03, 2024 @ 09:00 AM - 11:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Yuzhong Huang
Committee Members: Fred Morstatter (Chair), Yue Wang, Aiichiro Nakano, Antonio Ortega,
Title: Semantic Structure in Understanding and Generation of the 3D World
Abstract:
The ability to understand, generate, and modify 3D environments is foundational for applications such as virtual reality, autonomous driving, and generative AI tools. However, existing methods usually use non-semantic point clouds as their representation, which capture only geometric information without semantic context. This limitation creates a significant gap in both interpretability and performance when compared to methods that leverage semantic information. Moreover, non-semantic approaches often struggle to scale effectively as complexity increases, underscoring the importance of incorporating semantic structures to enhance scalability and adaptability.
This dissertation addresses these limitations by introducing methods that emphasize controllable semantic structures in 3D understanding, generation, and editing. First, to improve 3D scene understanding, we propose plane-aware techniques, such as planar priors and plane-splatting volume rendering, which provide explicit geometric and semantic representations. These methods enable more accurate and interpretable reconstructions compared to traditional point-cloud-based approaches. Second, for 3D content generation, we develop an orientation-conditioned diffusion model, which allows precise control over the alignment and orientation of generated objects, enhancing flexibility and user interaction. Third, to facilitate intuitive editing of 3D environments, we introduce a method for projecting text-guided 2D segmentation maps onto 3D models, bridging the gap between semantic understanding and user-driven modification.
These contributions collectively address the semantic and performance gaps in 3D reconstruction and generation, demonstrating that the integration of semantic information not only improves interpretability and precision but also enables models to scale more effectively for complex applications. By combining controllable semantic structures with geometric understanding, this dissertation advances the state-of-the-art in 3D vision and generation, paving the way for more scalable, interpretable, and interactive 3D workflows.
===============================
Time: Tuesday, December 3, 2024, 9:00 AM to 11:00 AM
Location: GCS | LL2 | SB-09
Zoom Link: https://usc.zoom.us/j/97579926743
Location: Ginsburg Hall (GCS) - SB-09
Audiences: Everyone Is Invited
Contact: Julia Mittenberg-Beirao
Event Link: https://usc.zoom.us/j/97579926743
-
PhD Dissertation Defense - Haiwei Chen
Mon, Dec 09, 2024 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Designing Neural Networks from the Perspective of Spatial Reasoning
Date: December 9, 2024
Time: 2:00 pm - 4:00 pm
Location: GFS 104
Committee: Yajie Zhao (Chair), Ramakant Nevatia, and Andrew Nealen
Abstract: All visual data, from image to CAD models, live in a 2D or 3D spatial domain. In order to understand and model the visual data, spatial reasoning has always been fundamental to computer vision algorithms. Naturally, the practice has been widely extended to the use of artificial neural networks built for visual analysis. The basic building blocks of a neural network - operators and representations - are means to learn spatial relationships and therefore are built with spatial properties. In this thesis, we present the designs of ``spatial-aware'' neural operators and representations in different application contexts, with a unique focus on how these design choices affect the spatial properties of the neural networks in a way that is beneficial for the tasks at hand. The first topic explored is the equivariance property, where a SE(3) equivariant convolutional network is designed for 3D pose estimation and scene registration. In this chapter, we show that the equivariant property of a convolutional neural network can be practically extended to higher dimensional space and proved highly effective for applications that are not only sensitive to translation, but also 3D rotations. The second topic explored is learning neural operators that approximate spatially continuous function in a pattern synthesis application context. In this chapter, we explore novel representations of periodic encoding and a continuous latent space for a generative network that is able to synthesize diverse, high-quality and continuous 2D and 3D patterns. The unique formulation allows the generative model to be at least 10 times faster and more memory efficient compared to previous efforts, and marked one of the earliest attempts to adopt the implicit network to the generative setting. The third topic explored is spatial awareness with regard to incomplete images, where a generative network model for image inpainting is designed based on restricting its receptive field. Combined with the generative transformer and the discrete latent codes, this novel paradigm demonstrates the effectiveness of separating analysis and synthesis in challenging image inpainting scenarios, as the resulted network model achieves state-of-the-art performance in both diversity and quality, when completing partial images with free-form holes occupying as large as 70\% of the image. I believe that the topics covered have contributed to a better understanding of neural operator and representation designs for both discriminative and generative learning in computer vision, from a perspective of identifying the effective ways of spatial reasoning for the targeted visual applications.Location: Grace Ford Salvatori Hall Of Letters, Arts & Sciences (GFS) - 104
Audiences: Everyone Is Invited
Contact: Ellecia Williams
