Depth Estimation Comparison
Compare 28 depth estimation models from 12 families side-by-side or with a slider.
Powered by the depth_estimation package.
⚡ Running on ZeroGPU — GPU resources are allocated automatically for inference.
Model A
Model B
Examples
Supported Models (28 variants · 12 families):
- Depth Anything v1 — ViT-S / ViT-B / ViT-L
- Depth Anything v2 — ViT-S / ViT-B / ViT-L
- Depth Anything v3 — Small / Base / Large / Giant / Nested Giant Large / Metric Large / Mono Large
- Intel ZoeDepth — metric depth fine-tuned on NYU + KITTI
- MiDaS — DPT-Large / DPT-Hybrid / BEiT-Large
- Apple DepthPro — sharp metric monocular depth in < 1 s
- Pixel-Perfect Depth — diffusion + MoGe metric alignment
- Marigold-DC — diffusion-based depth completion
- MoGe — v1 & v2 metric geometry (ViT-S/B/L + normal variants)
- OmniVGGT — metric depth ViT-L
- VGGT / VGGT-Commercial — metric depth from
facebook/VGGT-1B
Powered by depth_estimation · ZeroGPU manages GPU allocation automatically.