🏢 NVIDIA Research

Unlocking personalized image retrieval and segmentation, a novel approach uses pre-trained text-to-image diffusion models to surpass supervised methods, addressing limitations of existing self-supervi…

STAR, a novel unsupervised adaptation framework, drastically improves automatic speech recognition (ASR) robustness across diverse domains using only unlabeled data and outperforms existing self-train…

Large Spatial Model (LSM) achieves real-time semantic 3D reconstruction from just two unposed images, unifying multiple 3D vision tasks in a single feed-forward pass.

GaussianMarker: A novel uncertainty-aware watermarking method ensures robust copyright protection for 3D Gaussian Splatting assets, invisibly embedding messages into model parameters and extractable …

TRACKSTO4D: Fast & accurate 3D reconstruction from casual videos using 2D point tracks, drastically reducing runtime by up to 95% while matching state-of-the-art accuracy.

ESPACE: A novel LLM compression technique achieving 50% model size reduction with minimal accuracy loss by cleverly projecting activations onto principal components.

DistillNeRF: a self-supervised learning framework enabling accurate 3D scene reconstruction from sparse, single-frame images by cleverly distilling features from offline NeRFs and 2D foundation models…

CosAE: a novel autoencoder using learnable Fourier series achieves state-of-the-art image restoration by encoding frequency coefficients in its narrow bottleneck, preserving fine details even with ext…