Marzia Riso

2004, Route des Lucioles - 06902 Sophia Antipolis | Byron Building Y105 · marzia.riso@inria.fr

I am a Postdoctoral Researcher in the GraphDeco Group at Centre Inria d'Université Côte d'Azur. My current research focuses on direct manipulation and inverse procedural parameter estimation in a differentiable rendering scenario as well as geometry processing applications for porting vector graphics to the manifold domain.


Experience

Postdoctoral Researcher

Centre Inria d'Université Côte d'Azur
GraphDeco group
October 2024 - Current

Reseach Intern

Adobe Research

Six-months internship at Adobe Research Lab in Paris, France.

April 2023 - September 2023

Education

PhD student in Computer Graphics

Sapienza University of Rome
Advisor: Fabio Pellacini
Thesis: Enhancing Controllability in Procedural and Non-Procedural Asset Editing
Final grade: Cum laude
November 2020 - September 2024

MS degree in Computer Science

Sapienza University of Rome
Advisor: Maria De Marsico
Thesis: Adding interpretability to DNNs for breast cancer detection in Digital Pathology
Final grade: 110 cum laude
October 2018 - October 2020

BS Degree in Computer Science

Sapienza University of Rome
Advisor: Maria De Marsico
Thesis: Iris recognition via local operators combination
Final grade: 110 cum laude
September 2015 - October 2018

Publications

Direct Manipulation of Procedural Implicit Surfaces

Riso M., Michel È., Paris A., Deschaintre V., Gaillard M., Pellacini F.

This work proposes a method to directly manipulating of procedural implicit surfaces through mouse strokes in the viewport. We estimate an update of the procedural parameters of the implicit surface that matches the user intent thanks to the auto-differentiation of an augmented version of the implicit function.

ACM Transaction on Graphics (SIGGRAPH Asia)
DOI: TBA

2024

pEt: Direct Manipulation of Differentiable Vector Patterns

Riso M., Pellacini F.
pEt

pEt is a direct manipulation tool for editing the parameters of procedural vector patterns. Users selects some points in the viewport and expresses a transformation using a mouse movement. The parameters are optimized as the user performs the movement using a gradient-descent based approach.

Eurographics Symposium on Rendering (EGSR)
DOI: 10.2312/sr20231126

2023

BoolSurf: Boolean Operations on Surfaces

Riso M., Nazzaro G., Puppo E., Zhou Q., Jacobson A., Pellacini F.
Boolsurf

BoolSurf ports boolean set operations to the manifold settings. It computes boolean operations between geodesic shapes on surfaces of any genus and with open boundaries, supporting self-intersecting and complex designs.

ACM Transaction on Graphics (SIGGRAPH Asia) - video
DOI: 10.1145/3550454.3555466

2022

pOp: Parameter Optimization of Differentiable Vector Patterns

Riso M., Sforza D., Pellacini F.
pOp

pOp is a practical method for estimating the parameters of vector patterns, that are formed by collections of vector shapes arranged by an arbitrary procedural program. Program parameters are optimized by automatically differentiating the signed distance field of the patterns.

Computer Graphics Forum
DOI: 10.1111/cgf.14595

2022

Virtual Bowling! Launch as you all were there

De Marsico M., Panizzi E., Mattei F. R., Musolino A., Prandini M., Riso M., Sforza. D.
VirtualBowling

BowlingVR is an advanced Virtual Reality (VR) multiplayer game that provides a realistic User eXperience (UX) to the user. It provides realism in terms of the dynamics and physical context of a real bowling challenge as well as a remote and socially satisfying gameplay.

Proceedings of the ACM Int. Conf. on Advanced Visual Interfaces
DOI: 10.1145/3399715.3399848

2020

Honors & Awards

WomENcourage

ACM Celebration of Women in Computing
Scholarship winner and hackathon participant
2019

Skills

Programming Languages

Frameworks & Tools
Git | CLI | Jupyter | VSCode


Libraries
PyTorch | Tensorflow | Yocto/GL | OpenCV


Languages
Native: Italian | Fluent: English (B2 - First Certificate) | Spanish (A2)