Researchers have developed RD-SCL, a new framework for acoustic impedance imaging that addresses the challenge of scarce labeled data. This method integrates regularized deconvolution with semi-supervised cross-learning, utilizing a differentiable Tikhonov deconvolution operator to dynamically estimate the seismic wavelet during training. The framework enforces consistency between predictions on labeled and unlabeled data, outperforming existing methods on benchmarks like SEAM and Marmousi 2 with fewer parameters and lower computational cost. AI
IMPACT This framework offers a more efficient and physically consistent solution for subsurface analysis, potentially improving seismic exploration and resource discovery.
RANK_REASON The cluster describes a new research paper detailing a novel framework for a specific scientific imaging problem.
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
