Deformable Triangle Splatting: Flexible Primitives for Real-Time Radiance Field Rendering

Jul 24, 2026·
Oriol Jiménez-Ayguadé
Oriol Jiménez-Ayguadé
,
Antonio Agudo
· 0 min read
Abstract
Recent radiance field methods represent scenes with 2D primitives that offer surface alignment and efficient rasterization, from Gaussian disks to triangles, yet all rely on convex boundaries: curved and concave structures demand excessive primitives. We introduce Deformable Triangle Splatting, which augments each triangle with K control points per edge, each parameterized by a single learnable scalar displacement that shifts the boundary inward or outward, enabling non-convex shape representation while preserving the three base vertices that define the 3D plane. To render these non-convex primitives differentiably, we design a rasterization pipeline in the triangle’s barycentric coordinate space, ensuring view-consistent rendering. A winding number test determines whether each pixel lies inside the deformed primitive, and a window function controlled by two learnable parameters, sharpness and corner smoothness, together with a per-primitive scalar opacity, produces the smooth opacity transition from interior to boundary. Validation is done in a variety of real-world scenes, outperforming recent works based on non-volumetric primitives in terms of visual quality and versatility while still achieving competitive rendering efficiency.
Type
Publication
European Conference on Computer Vision (ECCV)
Status
Peer-reviewed
publications
Oriol Jiménez-Ayguadé
Authors
PhD Student in Computer Vision

I am a PhD student in Computer Vision at the Institute of Robotics and Industrial Informatics (IRI, CSIC–UPC) in Barcelona, advised by Antonio Agudo. My work sits at the intersection of computer vision and computer graphics, where I study how to turn ordinary images and video into faithful, editable 3D representations of the real world.

Broadly, I am interested in 3D scene understanding: reconstructing scenes, recognizing and segmenting the objects within them, and decomposing them into clean, manipulable parts that can be edited, relit, or reused. I work with mesh and splatting representations, multi-view segmentation, and generative models for 3D content, always with an eye toward connecting what a camera sees to the underlying structure of a scene.

Before starting my PhD I worked as an AI engineer and as a research assistant, on projects spanning 3D simulation, synthetic data generation, and interactive graphics. I also teach as a teaching assistant for the undergraduate Computer Graphics course at UPF. I enjoy building systems as much as studying them, and I am always happy to discuss research, 3D vision and graphics, or potential collaborations.