Showing 1–7 of 7 results for author: Seiskari, O

DN-Splatter: Depth and Normal Priors for Gaussian Splatting and Meshing

Authors: Matias Turkulainen, Xuqian Ren, Iaroslav Melekhov, Otto Seiskari, Esa Rahtu, Juho Kannala

Abstract: 3D Gaussian splatting, a novel differentiable rendering technique, has achieved state-of-the-art novel view synthesis results with high rendering speeds and relatively low training times. However, its performance on scenes commonly seen in indoor datasets is poor due to the lack of geometric constraints during optimization. We extend 3D Gaussian splatting with depth and normal cues to tackle chall… ▽ More 3D Gaussian splatting, a novel differentiable rendering technique, has achieved state-of-the-art novel view synthesis results with high rendering speeds and relatively low training times. However, its performance on scenes commonly seen in indoor datasets is poor due to the lack of geometric constraints during optimization. We extend 3D Gaussian splatting with depth and normal cues to tackle challenging indoor datasets and showcase techniques for efficient mesh extraction, an important downstream application. Specifically, we regularize the optimization procedure with depth information, enforce local smoothness of nearby Gaussians, and use the geometry of the 3D Gaussians supervised by normal cues to achieve better alignment with the true scene geometry. We improve depth estimation and novel view synthesis results over baselines and show how this simple yet effective regularization technique can be used to directly extract meshes from the Gaussian representation yielding more physically accurate reconstructions on indoor scenes. Our code will be released in https://github.com/maturk/dn-splatter. △ Less

Submitted 26 March, 2024; originally announced March 2024.

Gaussian Splatting on the Move: Blur and Rolling Shutter Compensation for Natural Camera Motion

Authors: Otto Seiskari, Jerry Ylilammi, Valtteri Kaatrasalo, Pekka Rantalankila, Matias Turkulainen, Juho Kannala, Arno Solin

Abstract: High-quality scene reconstruction and novel view synthesis based on Gaussian Splatting (3DGS) typically require steady, high-quality photographs, often impractical to capture with handheld cameras. We present a method that adapts to camera motion and allows high-quality scene reconstruction with handheld video data suffering from motion blur and rolling shutter distortion. Our approach is based on… ▽ More High-quality scene reconstruction and novel view synthesis based on Gaussian Splatting (3DGS) typically require steady, high-quality photographs, often impractical to capture with handheld cameras. We present a method that adapts to camera motion and allows high-quality scene reconstruction with handheld video data suffering from motion blur and rolling shutter distortion. Our approach is based on detailed modelling of the physical image formation process and utilizes velocities estimated using visual-inertial odometry (VIO). Camera poses are considered non-static during the exposure time of a single image frame and camera poses are further optimized in the reconstruction process. We formulate a differentiable rendering pipeline that leverages screen space approximation to efficiently incorporate rolling-shutter and motion blur effects into the 3DGS framework. Our results with both synthetic and real data demonstrate superior performance in mitigating camera motion over existing methods, thereby advancing 3DGS in naturalistic settings. △ Less

Submitted 24 May, 2024; v1 submitted 20 March, 2024; originally announced March 2024.

Comments: Source code available at https://github.com/SpectacularAI/3dgs-deblur

HybVIO: Pushing the Limits of Real-time Visual-inertial Odometry

Authors: Otto Seiskari, Pekka Rantalankila, Juho Kannala, Jerry Ylilammi, Esa Rahtu, Arno Solin

Abstract: We present HybVIO, a novel hybrid approach for combining filtering-based visual-inertial odometry (VIO) with optimization-based SLAM. The core of our method is highly robust, independent VIO with improved IMU bias modeling, outlier rejection, stationarity detection, and feature track selection, which is adjustable to run on embedded hardware. Long-term consistency is achieved with a loosely-couple… ▽ More We present HybVIO, a novel hybrid approach for combining filtering-based visual-inertial odometry (VIO) with optimization-based SLAM. The core of our method is highly robust, independent VIO with improved IMU bias modeling, outlier rejection, stationarity detection, and feature track selection, which is adjustable to run on embedded hardware. Long-term consistency is achieved with a loosely-coupled SLAM module. In academic benchmarks, our solution yields excellent performance in all categories, especially in the real-time use case, where we outperform the current state-of-the-art. We also demonstrate the feasibility of VIO for vehicular tracking on consumer-grade hardware using a custom dataset, and show good performance in comparison to current commercial VISLAM alternatives. An open-source implementation of the HybVIO method is available at https://github.com/SpectacularAI/HybVIO △ Less

Submitted 25 November, 2021; v1 submitted 22 June, 2021; originally announced June 2021.

Comments: 2022 IEEE Winter Conference on Applications of Computer Vision (WACV)

Point electrode problems in piecewise smooth plane domains

Authors: Otto Seiskari

Abstract: Conductivity equation is studied in piecewise smooth plane domains and with measure-valued current patterns (Neumann boundary values). This allows one to extend the recently introduced concept of bisweep data to piecewise smooth domains, which yields a new partial data result for Calderón inverse conductivity problem. It is also shown that bisweep data are (up to a constant scaling factor) the Sch… ▽ More Conductivity equation is studied in piecewise smooth plane domains and with measure-valued current patterns (Neumann boundary values). This allows one to extend the recently introduced concept of bisweep data to piecewise smooth domains, which yields a new partial data result for Calderón inverse conductivity problem. It is also shown that bisweep data are (up to a constant scaling factor) the Schwartz kernel of the relative Neumann-to-Dirichlet map. A numerical method for reconstructing the supports of inclusions from discrete bisweep data is also presented. △ Less

Submitted 21 May, 2013; v1 submitted 21 December, 2012; originally announced December 2012.

A note on analyticity properties of far field patterns

Authors: Roland Griesmaier, Nuutti Hyvönen, Otto Seiskari

Abstract: In scattering theory the far field pattern describes the directional dependence of a time-harmonic wave scattered by an obstacle or inhomogeneous medium, when observed sufficiently far away from these objects. Considering plane wave excitations, the far field pattern can be written as a function of two variables, namely the direction of propagation of the incident plane wave and the observation di… ▽ More In scattering theory the far field pattern describes the directional dependence of a time-harmonic wave scattered by an obstacle or inhomogeneous medium, when observed sufficiently far away from these objects. Considering plane wave excitations, the far field pattern can be written as a function of two variables, namely the direction of propagation of the incident plane wave and the observation direction, and it is well-known to be separately real analytic with respect to each of them. We show that the far field pattern is in fact a jointly real analytic function of these two variables. △ Less

Submitted 11 October, 2012; originally announced October 2012.

Comments: 8 pages

MSC Class: 35R30; 35Q60

GPU in Physics Computation: Case Geant4 Navigation

Authors: Otto Seiskari, Jukka Kommeri, Tapio Niemi

Abstract: General purpose computing on graphic processing units (GPU) is a potential method of speeding up scientific computation with low cost and high energy efficiency. We experimented with the particle physics simulation toolkit Geant4 used at CERN to benchmark its geometry navigation functionality on a GPU. The goal was to find out whether Geant4 physics simulations could benefit from GPU acceleration… ▽ More General purpose computing on graphic processing units (GPU) is a potential method of speeding up scientific computation with low cost and high energy efficiency. We experimented with the particle physics simulation toolkit Geant4 used at CERN to benchmark its geometry navigation functionality on a GPU. The goal was to find out whether Geant4 physics simulations could benefit from GPU acceleration and how difficult it is to modify Geant4 code to run in a GPU. We ported selected parts of Geant4 code to C99 & CUDA and implemented a simple gamma physics simulation utilizing this code to measure efficiency. The performance of the program was tested by running it on two different platforms: NVIDIA GeForce 470 GTX GPU and a 12-core AMD CPU system. Our conclusion was that GPUs can be a competitive alternate for multi-core computers but porting existing software in an efficient way is challenging. △ Less

Submitted 24 September, 2012; originally announced September 2012.

Point measurements for a Neumann-to-Dirichlet map and the Calderón problem in the plane

Authors: Nuutti Hyvönen, Petteri Piiroinen, Otto Seiskari

Abstract: This work considers properties of the Neumann-to-Dirichlet map for the conductivity equation under the assumption that the conductivity is identically one close to the boundary of the examined smooth, bounded and simply connected domain. It is demonstrated that the so-called bisweep data, i.e., the (relative) potential differences between two boundary points when delta currents of opposite signs a… ▽ More This work considers properties of the Neumann-to-Dirichlet map for the conductivity equation under the assumption that the conductivity is identically one close to the boundary of the examined smooth, bounded and simply connected domain. It is demonstrated that the so-called bisweep data, i.e., the (relative) potential differences between two boundary points when delta currents of opposite signs are applied at the very same points, uniquely determine the whole Neumann-to-Dirichlet map. In two dimensions, the bisweep data extend as a holomorphic function of two variables to some (interior) neighborhood of the product boundary. It follows that the whole Neumann-to-Dirichlet map is characterized by the derivatives of the bisweep data at an arbitrary point. On the diagonal of the product boundary, these derivatives can be given with the help of the derivatives of the (relative) boundary potentials at some fixed point caused by the distributional current densities supported at the same point, and thus such point measurements uniquely define the Neumann-to-Dirichlet map. This observation also leads to a new, truly local uniqueness result for the so-called Calderón inverse conductivity problem. △ Less

Submitted 2 April, 2012; originally announced April 2012.

MSC Class: 35R30; 35Q60