arXiv Science⌕ Search

arXiv subjects

Pak Hong Chui

Publications and source records attributed to Pak Hong Chui.

4 recordsLinked to original sources

Deep Bundle Adjustment

We present \textbf{DeepBA} (\textbf{Deep} \textbf{B}undle \textbf{A}djustment), a graph neural network (GNN) that recovers camera poses, intrinsics, and scene structure from 2D matches across an arbitrary number of views. Rather than solving a non-linear least-squares problem per scene, DeepBA amortizes bundle adjustment into a learned graph network that iteratively refines its estimate from geometric residuals. Where classical SfM computes its updates analytically and needs careful initialization, DeepBA learns them and converges from a cold start. It is trained with scale-invariant supervision (ground-truth rotation, translation direction, and intrinsics, plus reprojection self-consistency), so datasets with arbitrary metric units can be mixed without per-scene normalization. DeepBA is decoupled from the matcher. At 2.04M parameters, it outperforms prior learned SfM from matches on scene reconstruction and on most pose metrics, is competitive with classical SfM at moderate accuracy thresholds and overtakes it in wall time as the view count grows, and approaches monolithic image-to-geometry transformers such as VGGT ($\sim$1.2B parameters) with greater modularity. Project~page: https://cphjsn.github.io/deep-ba/

cs.CV↗

Track-Free Dense Structure from Motion

Classical Bundle Adjustment (BA) relies on precise metric initialization and known camera intrinsics, and traditional Structure-from-Motion (SfM) pipelines struggle to exploit modern dense matchers because rigid track-building heuristics break down under their inherent noise. We present Track-Free Dense Structure from Motion (TFD-SfM), which jointly optimizes camera extrinsics, focal lengths, and geometry from a strict cold start. Our central idea is to relax the rigidities of classical SfM into soft, probabilistic decisions under Negative Log-Likelihood (NLL) objectives. Instead of forcing matched pixels onto a single shared 3D point, we model each match as a 3D Gaussian whose variance is optimized jointly with the geometry. Unreliable matches take on large variance and contribute little to the loss. This smooths the optimization landscape, so the solver converges even from identity cameras and uniform depths. A kinematic relative-pose parameterization stabilizes the cold start, and a dual-hypothesis mechanism resolves the mirror-symmetry ambiguity on object-centric scenes. Across DTU, Mip-NeRF~360, ETH3D, and IMC2021, TFD-SfM is competitive with both classical and learning-based SfM baselines and leads on ETH3D, recovering accurate camera poses, intrinsics, and dense geometry from unordered, uncalibrated image collections. Project~page: https://cphjsn.github.io/tfd-sfm/

cs.CV↗

Topological superfluids with time-reversal symmetry from $s$-wave interaction in a bilayer system

Topological superconducting phases with time-reversal (TR) symmetry have been widely explored in recent years. However the involved unconventional pairings are generally implausible in realistic materials. Here we demonstrate via detailed self-consistent calculation that these topological phases with TR symmetry in DIII and BDI classes can be realized in a spin-orbit coupled bilayer system with only $s$-wave interaction. The bilayer freedom enables the definition of TR symmetry between the layers even in the presence of local Zeeman fields, which we propose to be realized using four laser beams. The gapped phase in DIII class is characterized by $\mathbb{Z}_2$, while all the gapless phases in these two classes are characterized by nontrivial winding numbers and are also manifested from the Majorana flat bands. We also unveil the intimate relation between TR symmetry and mirror symmetry due to phase locking effect between the two layers, which harbors the mirror symmetry protected topological phases. We finally demonstrate that these phases will not be spoiled by interlayer pairings.

cond-mat.quant-gas↗

Unconventional pairings of spin-orbit coupled attractive degenerate Fermi gas in a one dimensional optical lattice

Understanding novel pairings in attractive degenerate Fermi gases is crucial for exploring rich superfluid physics. In this report, we reveal unconventional pairings induced by spin-orbit coupling (SOC) in a one-dimensional optical lattice, using a state-of-the-art density-matrix renormalization group method. When both bands are partially occupied, we find a strong competition between the interband Fulde-Ferrell-Larkin-Ovchinnikov (FFLO) and intraband Bardeen-Cooper-Schrieffer (BCS) pairings. In particular, for the weak and moderate SOC strengths, these two pairings can coexist, giving rise to a new phase called the FFLO-BCS phase, which exhibits a unique three-peak structure in pairing momentum distribution. For the strong SOC strength, the intraband BCS pairing always dominates in the whole parameter regime, including the half filling. We figure out the whole phase diagrams as functions of filling factor, SOC strength, and Zeeman field. Our results are qualitatively different from recent mean-field predictions. Finally, we address that our predictions could be observed in a weaker trapped potential.

cond-mat.quant-gas↗