Lior David | Searcharxiv

arXiv2026

Ego-1K -- A Large-Scale Multiview Video Dataset for Egocentric Vision

We present Ego-1K, a large-scale collection of time-synchronized egocentric multiview videos designed to advance neural 3D video synthesis and dynamic scene understanding. The dataset contains nearly 1,000 short egocentric videos captured with a custom rig with 12 synchronized cameras surrounding a 4-camera VR headset worn by the user. Scene content focuses on hand motions and hand-object interactions in different settings. We describe rig design, data processing, and calibration. Our dataset enables new ways to benchmark egocentric scene reconstruction methods, an important research area as smart glasses with multiple cameras become omnipresent. Our experiments demonstrate that our dataset presents unique challenges for existing 3D and 4D novel view synthesis methods due to large disparities and image motion caused by close dynamic objects and rig egomotion. Our dataset supports future research in this challenging domain. It is available at https://huggingface.co/datasets/facebook/ego-1k.

cs.CV↗

arXiv2013

Efficient Parallel Computation of the Estimated Covariance Matrix

Computation of a signal's estimated covariance matrix is an important building block in signal processing, e.g., for spectral estimation. Each matrix element is a sum of products of elements in the input matrix taken over a sliding window. Any given product contributes to multiple output elements, thereby complicating parallelization. We present a novel algorithm that attains very high parallelism without repeating multiplications or requiring inter-core synchronization. Key to this is the assignment to each core of distinct diagonal segments of the output matrix, selected such that no multiplications need to be repeated yet only one core writes to any given output-matrix element, and exploitation of a shared memory (including L1 cache) that obviates the need for a corresponding awkward partitioning of the memory among cores. Implementation on Plurality's HyperCore shared-memory many-core architecture demonstrates linear speedup of up to 64 cores and speedups of ~85X for 128 cores. On an x86 system we demonstrate that the new algorithm has consider parallel speedups but also show that a sequential implementation of the new algorithm outperforms the parallel implementation of the baseline approach. On a quad-core x86 system, the new algorithm is 20X faster than sequential baseline and 5X than parallel implementation of the baseline.

cs.DS↗