Puzzles: Unbounded Video-Depth Augmentation for Scalable End-to-End 3D Reconstruction
This addresses data scarcity for scalable end-to-end 3D reconstruction in computer vision, offering an incremental improvement by enhancing existing pipelines without architectural changes.
The paper tackles the limited diversity and scale of training data for multi-view 3D reconstruction by introducing Puzzles, a data augmentation strategy that synthesizes unbounded posed video-depth data from single images or clips, resulting in models trained on only 10% of original data achieving comparable accuracy to full-dataset training.
Multi-view 3D reconstruction remains a core challenge in computer vision. Recent methods, such as DUST3R and its successors, directly regress pointmaps from image pairs without relying on known scene geometry or camera parameters. However, the performance of these models is constrained by the diversity and scale of available training data. In this work, we introduce Puzzles, a data augmentation strategy that synthesizes an unbounded volume of high-quality posed video-depth data from a single image or video clip. By simulating diverse camera trajectories and realistic scene geometry through targeted image transformations, Puzzles significantly enhances data variety. Extensive experiments show that integrating Puzzles into existing video-based 3D reconstruction pipelines consistently boosts performance without modifying the underlying network architecture. Notably, models trained on only ten percent of the original data augmented with Puzzles still achieve accuracy comparable to those trained on the full dataset. Code is available at https://jiahao-ma.github.io/puzzles/.