Distill3R: A Pipeline for Democratizing 3D Foundation Models on Commodity Hardware

TL;DR

Distill3R introduces a method to distill large 3D foundation models into compact, trainable models on standard hardware, making advanced 3D reconstruction accessible to smaller labs.

Contribution

The paper presents a novel distillation framework with offline caching and confidence-aware loss, enabling training of 3D models on commodity hardware, reducing size and inference time significantly.

Findings

Student model achieves 9x parameter reduction

Inference speed is 5x faster than teacher

Training completed in under 3 days on a single workstation

Abstract

While multi-view 3D reconstruction has shifted toward large-scale foundation models capable of inferring globally consistent geometry, their reliance on massive computational clusters for training has created a significant barrier to entry for most academic laboratories. To bridge this compute divide, we introduce Distill3R, a framework designed to distill the geometric reasoning of 3D foundation models into compact students fully trainable on a single workstation. Our methodology centers on two primary innovations: (1) an offline caching pipeline that decouples heavy teacher inference from the training loop through compressed supervision signals, and (2) a confidence-aware distillation loss that leverages teacher uncertainty to enable training on commodity hardware. We propose a 72M-parameter student model which achieves a 9x reduction in parameters and a 5x inference speedup compared to its 650M-parameter teacher. The student is fully trainable in under 3 days on a single workstation, whereas its teacher requires massive GPU clusters for up to a week. We demonstrate that the student preserves the structural consistency and qualitative geometric understanding required for functional 3D awareness. By providing a reproducible, single-workstation training recipe, Distill3R serves as an exploratory entry point for democratized 3D vision research and efficient edge deployment. This work is not intended to compete with state-of-the-art foundation models, but to provide an accessible research baseline for laboratories without access to large-scale compute to train and specialize models on their own domain-specific data at minimal cost.