Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

DefVINS: Visual-Inertial Odometry for Deformable Scenes

About

Deformable scenes violate the rigidity assumptions underpinning classical visual-inertial odometry (VIO), often leading to over-fitting to local non-rigid motion or severe drift when deformation dominates visual parallax. We introduce DefVINS, a visual-inertial odometry framework that explicitly separates a rigid, IMU-anchored state from a non--rigid warp represented by an embedded deformation graph. The system is initialized using a standard VIO procedure that fixes gravity, velocity, and IMU biases, after which non-rigid degrees of freedom are activated progressively as the estimation becomes well conditioned. An observability analysis is included to characterize how inertial measurements constrain the rigid motion and render otherwise unobservable modes identifiable in the presence of deformation. This analysis motivates the use of IMU anchoring and informs a conditioning-based activation strategy that prevents ill-posed updates under poor excitation. Ablation studies demonstrate the benefits of combining inertial constraints with observability-aware deformation activation, resulting in improved robustness under non-rigid environments.

Samuel Cerezo, Javier Civera• 2026

Related benchmarks

TaskDatasetResultRank
Visual-Inertial OdometryReal Deformable Sequences Medium Deformation R2
ATE RMSE (mm)10.2
5
Visual-Inertial OdometryReal Deformable Sequences Medium Deformation R3
ATE RMSE10.8
5
Visual-Inertial OdometryReal Deformable Sequences Medium Deformation R4
ATE RMSE (mm)11.4
5
Visual-Inertial OdometryReal Deformable Sequences R5 (High Deformation)
ATE RMSE (mm)15.6
5
Visual-Inertial OdometryReal Deformable Sequences High Deformation R6
ATE RMSE (mm)19.8
5
Visual-Inertial SLAMDrunkard's Dataset Medium deformation - L1
ATE RMSE (mm)9.4
5
Visual-Inertial SLAMDrunkard's Dataset Hard deformation - L2
ATE RMSE (mm)14.3
5
Visual-Inertial SLAMDrunkard's Dataset Extreme deformation - L3
ATE RMSE (mm)19.6
5
Visual-Inertial OdometryReal Deformable Sequences Low Deformation R0
ATE RMSE8.1
5
Visual-Inertial OdometryReal Deformable Sequences Low Deformation R1
ATE RMSE (mm)9
5
Showing 10 of 11 rows

Other info

Follow for update