Depth Anything
A foundation model for monocular depth estimation that infers relative depth from a single image. It substantially improves relative depth quality; however, its output consists of scale-free relative values rather than absolute metric measurements.
Timeline (last 90 days)
Open-source models such as Depth Anything and MiDaS have significantly lowered the barrier to using monocular depth estimation.
Representative open-source models for monocular depth estimation include MiDaS and Depth Anything, which can run locally on consumer-grade GPUs.
现代 AI 视频生成模型基于变分自编码器将高维像素图像压缩至低维潜向量表示，在潜空间中寻找过渡路径而非像素级线性插值
All Facts (4)
Open-source models such as Depth Anything and MiDaS have significantly lowered the barrier to using monocular depth estimation.
50%UnverifiedRepresentative open-source models for monocular depth estimation include MiDaS and Depth Anything, which can run locally on consumer-grade GPUs.
50%Unverified现代 AI 视频生成模型基于变分自编码器将高维像素图像压缩至低维潜向量表示，在潜空间中寻找过渡路径而非像素级线性插值
50%Unverified单目深度估计的代表性模型包括Intel的MiDaS、Meta的DPT以及Depth Anything系列
50%