UniK3D (fastforks)

UniK3D (Piccinelli et al. 2025): monocular metric 3-D for any camera model; a DINOv2 ViT-L encoder and a camera-prompted decoder predict a spherical-harmonic ray model and per-pixel radii (depth, 3-D points, confidence), or use given pinhole intrinsics.

This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: https://github.com/kleinicke/fastforks. Kernels for the kernels backend: https://huggingface.co/fastforks/kernels. Original code: https://github.com/lpiccinelli-eth/UniK3D at 29f7862.

Author weight repos: https://huggingface.co/lpiccinelli/unik3d-vitl.

Task: depth. Tier: experimental. Outputs: depth, points, confidence. Backends: portable, kernels.

Input: one RGB image. Output: a depth or inverse-depth map. kind (a card field, result.kind) says what the numbers mean: metric = metres; scale_invariant, affine_invariant and affine_invariant_inverse are relative (unknown scale or shift, not metres). Metric models that use the focal length take model(image, intrinsics=K) (K [3, 3], pixels of the input): metric3d-vit-large and da3metric-large need it for correct metres (da3metric-large: metres = output x focal / 300; without intrinsics= a 300 px focal length is assumed, wrong for a real camera). unidepth-v2-vitl14, unik3d-vitl, depth-pro and MoGe estimate the camera themselves (given intrinsics= replace fx); zoedepth-nk and the relative models (da3mono-large too) take none.

Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in card.toml.

variant	weights (author link)	SHA-256	weights licence	commercial use
unik3d-vitl@1	model.safetensors	890be79699dd	CC-BY-NC-SA-4.0 (assumed from code)	no

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: CC-BY-NC-SA-4.0. Files: LICENSE.

Code licence: CC-BY-NC-SA-4.0; the authors do not state a separate weights licence.

Training data: Mixture of public metric-depth datasets (pinhole, fisheye and panoramic cameras) per the UniK3D paper (arXiv 2503.16591).

Licence evidence: https://github.com/lpiccinelli-eth/UniK3D/blob/29f7862f3c3ee79c53ef0153dfc901ce511df575/LICENSE

Input size

any H x W >= 8, resized or padded inside the model (ViT patch 14) Tested up to (portable, fp16, RTX 5060 Ti 16 GB): unik3d-vitl 3000x4000 (2.2 GiB peak allocated, 2.4 reserved)

Speed and memory

Depth, Middlebury 2014 half, 994x1482

model	time (ms): original fp32	original	portable	kernels	kernels speed-up: vs original	vs portable	peak memory (MiB, allocated / reserved): original	portable	kernels
unik3d-vitl	133.7	133.9	107.77	62.92	2.1x	1.7x	3307 / 3570	1219 / 1492	1166 / 1266

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
unik3d-vitl	7.28	4.16	0.84†	0.57	0.14	0.07

torch.compile and TensorRT

model	input	portable ms	compile ms	compile s (cold)	TensorRT fp16 ms	ONNX + build s	kernels ms	peak MiB (alloc / reserved): portable	compile	TensorRT (device)
unik3d-vitl	994x1482	107.77	94.4	32.8	fails: TensorRT: antialiased resize not supported	–	62.92	1219 / 1492	1258 / 1312	–

Fixed-size speed and peak memory, all models

model	mode	640x480	1280x720	1920x1080	12 MP 4000x3000
unik3d-vitl	portable	115.1 ms / 2101 / 2474	124.2 ms / 2153 / 2570	124.8 ms / 2166 / 2582	128.7 ms / 2271 / 2486
unik3d-vitl	kernels	57.9 ms / 1076 / 1202	62.6 ms / 1101 / 1272	63.2 ms / 1115 / 1240	65.9 ms / 1226 / 1512

Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).

Hardware, protocol and caveats: benchmarks (RTX 5060 Ti 16 GB, sm_120). Original = the authors' code at the precision of portable (fp16, or (fp32) where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's.

Accuracy

Parity with the original code

Every variant is checked against golden outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 1 % and p99 <= 2 % of the reference's median |value| (absolute error of the output depth).

Usage

from fastforks import depth

# non-commercial weights: run with FASTFORKS_NONCOMMERCIAL=1 in the environment
model = depth.load("unik3d-vitl")  # mode="auto": the preferred backend from recorded measurements
result = model("room.jpg")
result.depth  # [H, W] float32 tensor; result.kind says what the maps mean
# models that take a camera: model("room.jpg", intrinsics=K), K [3, 3] in pixels
result.backend  # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')

Install with pip install fastforks (see https://github.com/kleinicke/fastforks). Other variants: unik3d-vitl.

Kernels backend

depth.load("unik3d-vitl", mode="kernels") runs our CUDA kernels (fastbb, dpt_fused). It needs a prebuilt kernel for your GPU architecture (downloaded from https://huggingface.co/fastforks/kernels on first use) and a recent NVIDIA driver; if there is none it raises BackendUnavailableError with the reason and never falls back silently. mode="auto" picks kernels when it can run and logs the choice.