arXiv · Statistics Machine Learning· Alessio Spagnoletti, Charlesquin Kemajou Mbakam, Jonathan Spence, Andr\'es Almansa, Marcelo Pereyra·· 3 小时前AI 评分40
BAM (Bayesian Anything Model)：面向生成式计算成像的基础模型
BAM! Bayesian Anything Model: a foundation model for generative computational imaging
AI 导读
BAM 是一个仅 36M 参数的轻量基础模型，将 operator-conditioned Reconstruct Anything Model（RAM）骨干升级为条件流映射，在推理时指定仪器物理机制。
来源：arXiv · Statistics Machine Learning · arxiv.org