S23DR 2026 Winning Solution
For researchers in 3D reconstruction, this work provides a top-performing method on a specific benchmark, but is incremental as it combines existing techniques (DiT, Perceiver, flow matching) without introducing a new paradigm.
The winning solution to the S23DR 2026 challenge for structured 3D wireframe reconstruction achieves first place with HSS=0.654 by treating vertices as a conditional set and denoising with flow-matching DiT.
This text presents the winning solution to the S23DR 2026 challenge for structured 3D wireframe reconstruction from sparse SfM, fitted depth, and semantic segmentations. The method treats vertices as a conditional set and denoises 64 vertex tokens with a flow-matching DiT conditioned on Perceiver-style scene tokens. A global pass predicts the coarse structure, a hull-cropped second pass refines it, and a small multi-sample consensus step keeps the stochastic sampler well behaved. The final system ranked first on the private leaderboard, achievingHSS = 0.654.