Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a novel global Structure-from-Motion framework that utilizes a permutation-equivariant, edge-conditioned graph neural network to estimate camera poses from noisy pairwise relative poses. By training without ground-truth supervision and focusing on a relative-pose consistency objective, the method achieves globally consistent camera extrinsics, enabling efficient and scalable processing of over a thousand images. Experimental results on multiple datasets show that the proposed approach outperforms existing deep track-centric methods in rotation and translation accuracy while maintaining competitive performance against classical pipelines, all with significantly faster processing times.
Achieving globally consistent camera poses from noisy data without ground-truth supervision could revolutionize 3D reconstruction workflows.
Camera pose estimation is a key step in 3D reconstruction and view-synthesis pipelines. We present a deep, global Structure-from-Motion framework based on learned view-graph aggregation. Our method employs a permutation-equivariant, edge-conditioned graph neural network that takes noisy pairwise relative poses as input and outputs globally consistent camera extrinsics. The network is trained without ground-truth supervision, relying solely on a relative-pose consistency objective. This is followed by 3D point triangulation and robust bundle adjustment. Our approach is efficient, scalable to more than a thousand images, and robust to graph density. We evaluate our method on MegaDepth, 1DSfM, Strecha, and BlendedMVS. These experiments demonstrate that our method achieves superior rotation and translation accuracy compared to deep track-centric methods while registering more images across many scenes, and competitive results compared to state-of-the-art classical pipelines, while being much faster.