Media Depth
Monocular depth estimation with open-source + commercial-safe models: Depth-Anything v2 (Apache 2.0, TikTok/ByteDance 2024, current SOTA small/base/large) via HuggingFace transformers, MiDaS v3.1 (MIT, Intel ISL classic with DPT/BEiT/Swin backbones). Extract depth maps from a single image or vide...
