arXiv subjects

Dmitrii Pozdeev

Publications and source records attributed to Dmitrii Pozdeev.

4 recordsLinked to original sources

arXiv2026

Generative Shape Reconstruction with Geometry-Guided Langevin Dynamics

Reconstructing complete 3D shapes from incomplete or noisy observations is a fundamentally ill-posed problem that requires balancing measurement consistency with shape plausibility. Existing methods for shape reconstruction can achieve strong geometric fidelity in ideal conditions but fail under realistic conditions with incomplete measurements or noise. At the same time, recent generative models for 3D shapes can synthesize highly realistic and detailed shapes but fail to be consistent with observed measurements. In this work, we introduce GG-Langevin: Geometry-Guided Langevin dynamics, a probabilistic approach that unifies these complementary perspectives. By traversing the trajectories of Langevin dynamics induced by a diffusion model, while preserving measurement consistency at every step, we generatively reconstruct shapes that fit both the measurements and the data-informed prior. We demonstrate through extensive experiments that GG-Langevin achieves higher geometric accuracy and greater robustness to missing data than existing methods for surface reconstruction.

cs.CV↗

arXiv2025

Densemarks: Learning Canonical Embeddings for Human Heads Images via Point Tracks

We propose DenseMarks - a new learned representation for human heads, enabling high-quality dense correspondences of human head images. For a 2D image of a human head, a Vision Transformer network predicts a 3D embedding for each pixel, which corresponds to a location in a 3D canonical unit cube. In order to train our network, we collect a dataset of pairwise point matches, estimated by a state-of-the-art point tracker over a collection of diverse in-the-wild talking heads videos, and guide the mapping via a contrastive loss, encouraging matched points to have close embeddings. We further employ multi-task learning with face landmarks and segmentation constraints, as well as imposing spatial continuity of embeddings through latent cube features, which results in an interpretable and queryable canonical space. The representation can be used for finding common semantic parts, face/head tracking, and stereo reconstruction. Due to the strong supervision, our method is robust to pose variations and covers the entire head, including hair. Additionally, the canonical space bottleneck makes sure the obtained representations are consistent across diverse poses and individuals. We demonstrate state-of-the-art results in geometry-aware point matching and monocular head tracking with 3D Morphable Models. The code and the model checkpoint will be made available to the public.

cs.CV↗

arXiv2024

Ventilated noise-insulating metamaterials inspired by sonic black holes

Acoustic black holes represent a special class of metastructures allowing efficient absorption based on the slow sound principle. The decrease of the wave speed is associated with the spatial variation of acoustic impedance, while the absorption properties are linked to thermoviscous losses induced by the local resonances of the structure. While most of the developments in the field of sonic black holes are dedicated to one-dimensional structures, the current study is concerned with their two-dimensional counterparts. It is shown that the change of the dimensionality results in the change of noise insulation mechanism, which relies on the opening of band-gaps rather then thermoviscous losses. The formation of band-gaps is associated with the strong coupling between the resonators constituting the considered structures. Numerically and experimentally it is shown than the structure is characterized by broad stop-bands in transmission spectra, while the air flow propagation is still allowed. In particular, a realistic application scenario is considered, in which the acoustic noise and the air flow are generated by a fan embedded into a ventilation duct. The obtained results pave the way towards the development of next-level ventilated metamaterials for efficient noise control.

physics.app-ph↗

arXiv2024

To Stay or Not to Stay in the Pre-train Basin: Insights on Ensembling in Transfer Learning

Transfer learning and ensembling are two popular techniques for improving the performance and robustness of neural networks. Due to the high cost of pre-training, ensembles of models fine-tuned from a single pre-trained checkpoint are often used in practice. Such models end up in the same basin of the loss landscape, which we call the pre-train basin, and thus have limited diversity. In this work, we show that ensembles trained from a single pre-trained checkpoint may be improved by better exploring the pre-train basin, however, leaving the basin results in losing the benefits of transfer learning and in degradation of the ensemble quality. Based on the analysis of existing exploration methods, we propose a more effective modification of the Snapshot Ensembles (SSE) for transfer learning setup, StarSSE, which results in stronger ensembles and uniform model soups.

cs.LG↗