Zhenwei Shi

Just three satellite views are now enough for high-fidelity 3D radiance field reconstruction, beating previous sparse-view baselines by a massive 6.83 dB PSNR.

Single-model 3D Gaussian Splatting can match the calibrated uncertainty bounds of a 10-model ensemble at 280 FPS simply by factorizing spatial rendering geometry from view-dependent difficulty under a conformal framework.

Achieving linear-time efficiency in remote sensing instance segmentation could revolutionize how we deploy vision models in resource-constrained environments.

LEVIRDet-159 not only sets a new standard in dataset scale but also enables a foundation model that generalizes effectively across diverse remote sensing contexts.

SSMs are revolutionizing remote sensing by efficiently tackling complex data challenges, yet many researchers remain unaware of their full potential.

Scale-Frustum Representations offer a surprisingly effective way to model objects at drastically different scales in ultra-wide area remote sensing images, boosting segmentation accuracy and convergence speed.

Achieve state-of-the-art cloud detection in remote sensing imagery with a novel Mamba-based architecture that's both more accurate and computationally efficient than Transformers.

A breast pathology foundation model, BRIGHT, beats generalist models at specialist tasks, suggesting organ-specific pretraining is a scalable path to expert-level performance.