Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective

TL;DR

This paper surveys recent feed-forward 3D scene modeling methods, emphasizing model design strategies across diverse geometric representations, and proposes a taxonomy to organize research directions and applications.

Contribution

It introduces a novel, output-agnostic taxonomy of model design strategies for feed-forward 3D reconstruction and provides a comprehensive review of benchmarks, datasets, and applications.

Findings

Shared architectural patterns across diverse geometric representations

Five key research problems: feature enhancement, geometry awareness, efficiency, augmentation, temporal models

Discussion of future challenges like scalability and evaluation standards

Abstract

Reconstructing 3D representations from 2D inputs is a fundamental task in computer vision and graphics, serving as a cornerstone for understanding and interacting with the physical world. While traditional methods achieve high fidelity, they are limited by slow per-scene optimization or category-specific training, which hinders their practical deployment and scalability. Hence, generalizable feed-forward 3D reconstruction has witnessed rapid development in recent years. By learning a model that maps images directly to 3D representations in a single forward pass, these methods enable efficient reconstruction and robust cross-scene generalization. Our survey is motivated by a critical observation: despite the diverse geometric output representations, ranging from implicit fields to explicit primitives, existing feed-forward approaches share similar high-level architectural patterns, such as image feature extraction backbones, multi-view information fusion mechanisms, and geometry-aware design principles. Consequently, we abstract away from these representation differences and instead focus on model design, proposing a novel taxonomy centered on model design strategies that are agnostic to the output format. Our proposed taxonomy organizes the research directions into five key problems that drive recent research development: feature enhancement, geometry awareness, model efficiency, augmentation strategies and temporal-aware models. To support this taxonomy with empirical grounding and standardized evaluation, we further comprehensively review related benchmarks and datasets, and extensively discuss and categorize real-world applications based on feed-forward 3D models. Finally, we outline future directions to address open challenges such as scalability, evaluation standards, and world modeling.