marigold / README.md
Flor's picture
card: update from fastforks 1cbf1671
4897bd4 verified
|
Raw History Blame Contribute Delete
9.02 kB
metadata
library_name: fastforks
pipeline_tag: depth-estimation
license: openrail++
tags:
  - depth
  - mono-depth
  - monocular-depth
  - cuda
  - pytorch
  - fastforks
  - fastforks-card
base_model:
  - prs-eth/marigold-depth-v1-1

Marigold (fastforks)

Marigold (Ke et al. 2023): affine-invariant monocular depth by diffusion; a Stable Diffusion 2 UNet fine-tuned to denoise a depth latent conditioned on the image latent (depth v1-1, few DDIM steps).

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/prs-eth/Marigold at 2bfbdea.

Author weight repos: https://huggingface.co/prs-eth/marigold-depth-v1-1.

Task: depth. Tier: stable. Outputs: depth. 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, affine_invariant_log (log depth) 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. Normals-only models (kind none) return result.normals [B, H, W, 3] (unit vectors, camera frame: x right, y down, z forward).

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
marigold-depth-v1-1@1 unet/diffusion_pytorch_model.safetensors
vae/diffusion_pytorch_model.safetensors
text_encoder/model.safetensors
ad9342d79bb5
a1d993488569
cce6febb0b6d
LicenseRef-CreativeML-OpenRAIL-PlusPlus-M yes

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: Apache-2.0. Licence files: LICENSE, LICENSE-MODEL, LICENSE-huggingface.

Training data: Stable Diffusion 2 (LAION-5B subsets) fine-tuned on the synthetic Hypersim and Virtual KITTI 2 datasets per the Marigold papers (arXiv 2312.02145, 2505.09358).

  • Licence note: The weights use the OpenRAIL++-M licence: commercial use is allowed but with use restrictions that must be passed on; the training data terms are not checked.

Licence evidence: https://huggingface.co/prs-eth/marigold-depth-v1-1/blob/9571e7123e258cf052b4e54241f17971c290e9a8/README.md (License: CreativeML Open RAIL++-M)

Input size

any H x W >= 8, resized or padded inside the model Tested up to (portable, fp16, RTX 5060 Ti 16 GB): marigold-depth-v1-1 3000x4000 (3.0 GiB peak allocated, 3.8 reserved)

Speed and memory

Depth, synthetic

model input time (ms): original fp32 original portable kernels kernels speed-up: vs original vs portable peak memory (MiB, allocated / reserved): original portable kernels
marigold-depth-v1-1 994x1482 – – 410 261 – 1.6x – 2776 / 2938 2343 / 2500

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
marigold-depth-v1-1 963 439 – – – – 3414 / 3580 – –

Startup and first call

model load (s): original portable kernels first call (s): original portable kernels
marigold-depth-v1-1 1.32 11.04 (518x770) 2.49 (518x770) 0.60 0.89 (518x770) 0.50 (518x770)

torch.compile and TensorRT

model input time: portable compile TensorRT fp16 kernels preparation: compile (cold first call) ONNX + engine build peak MiB (alloc / reserved): portable compile kernels TensorRT (engine + torch)
marigold-depth-v1-1 994x1482 410 ms 342 ms 172 ms 261 ms 127 s 285 s 2776 / 2938 2341 / 2418 2343 / 2500 4263 / 4318

Fixed-size speed and peak memory, all models

model mode 480x640 720x1280 1080x1920 12 MP 3000x4000
marigold-depth-v1-1 portable 455.3 ms / 2898 / 3650 351.7 ms / 2648 / 3306 351.8 ms / 2661 / 3378 458.1 ms / 3033 / 3844
marigold-depth-v1-1 kernels 433.0 ms / 2636 / 3464 324.3 ms / 2452 / 3070 324.9 ms / 2465 / 3142 435.6 ms / 2770 / 3552

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 original code named above 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

On public datasets

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Original = the original code named under Details, same checkpoint and protocol.

variant dataset original fp32 original fp16 portable kernels note
marigold-depth-v1-1 NYUv2 test, AbsRel / delta1 0.0561 / 0.9703 0.0561 / 0.9703 0.0561 / 0.9703 0.0561 / 0.9703
marigold-depth-v1-1 iBims-1, AbsRel / delta1 0.0499 / 0.9721 0.0499 / 0.9721 0.0499 / 0.9722 0.0499 / 0.9721
marigold-depth-v1-1 ETH3D two-view train left, AbsRel / delta1 0.1141 / 0.8715 0.1141 / 0.8715 0.1141 / 0.8715 0.1141 / 0.8716 possibly in the training data
marigold-depth-v1-1 Middlebury 2014 trainingH left, AbsRel / delta1 0.0413 / 0.9899 0.0413 / 0.9899 0.0413 / 0.9899 0.0413 / 0.9899 possibly in the training data

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 491e248a / 939adfe8, 2026-11-06 to 2026-11-11.

Parity with the original code

Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.0001 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 1 % and p99 <= 3 % of the reference's median |value| (absolute error of the output depth).

Limitations

  • Diffusion model: the seeded noise comes from a CPU generator and differs from the original code's CUDA generator for the same seed.
  • A traced or exported module reuses the noise of its traced shape.
  • Kernels time is dominated by unfused elementwise kernels (GroupNorm / SiLU / adds).

Usage

from fastforks import depth

model = depth.load("marigold-depth-v1-1")  # 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: marigold-depth-v1-1.

Kernels backend

depth.load("marigold-depth-v1-1", mode="kernels") runs our CUDA kernels (cudnn_hacc). 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.