Models · Monocular depth

MoGe

Microsoft's MoGe: depth, 3D points, normals and focal length from one image.

stable Fork of microsoft/MoGe@74fbce0 · backends: portable, kernels

Checkpoint

Paper 2025 Added in 0.1.0 Evaluated with 0.2.0

Kernels, RTX 5060 Ti 16 GB 33.3 ms 994x1482, fp16
vs the faster original 2.5× original fp16: 82.6 ms
Peak memory 932 MiB original: 1,193 MiB
Accuracy vs original fp32 ±0.0000 kernels AbsRel, NYUv2 test (0.0484 original)

Use it

pip install fastforks

Lean install with only what this model needs: pip install fastforks-core

from fastforks import depth

model = depth.load("moge-2-vits-normal")  # kernels on a supported GPU, else portable PyTorch
result = model(image)
result.depth  # also: points, intrinsics, mask, normals
result.backend  # what ran: mode, device, dtype, kernels
fastforks bench moge-2-vits-normal

Times the model on this machine. load(…, mode="portable") or mode="kernels" picks a backend; usage guide.

Speed and memory

Every backend on one GPU (RTX 5060 Ti 16 GB) and the same input, the authors’ code included. Kernels runs fp16. Portable runs fp16 on CUDA.

BackenddtypeTime per call (ms)vs original fp16Peak memory
allocated / reserved
Load (s)First call (s)
Original fp32fp32 1620.51×1,491 MiB/1,628 MiB0.40.22
Original fp16fp16 82.6–1,193 MiB/1,268 MiB0.91.50
Portable fp32fp32not measured ––––
Portablefp16 61.61.3×1,007 MiB/1,172 MiB1.20.06
Kernelsfp16 33.32.5×932 MiB/1,048 MiB0.20.03
Portable + torch.compile 43.91.9×836 MiB/992 MiB124 (build)–
TensorRT fp16fp16 24.53.4×620 MiBenginenot comparable131 (build)–

Measured on RTX 5060 Ti 16 GB, torch 2.14.1+cu130, 2026-11-05 to 2026-11-15. First call, 3 warm-up, peak reset, median of 3 rounds x median of 5. Ratios divide two times of this table (same GPU, same input). Original fp16 is the authors’ own precision option, else torch.autocast. torch.compile and TensorRT: build time instead of load time. TensorRT memory (engine, not comparable) is the GPU memory the engine took, read from the driver because TensorRT allocates outside PyTorch, so it can't be compared with the PyTorch peaks of the other rows.

Other devices

DeviceBackenddtypems / callPeak memoryOutput vs referenceInput
Apple M1 Pro 16 GB MPS, torch 2.14.1, 2026-11-14Originalfp32516 2,849 MiB processmatches994x1482
Portablefp16857 1,679 MiB processmatches994x1482
Portablefp32361 2,634 MiB processmatches994x1482
Kernelsfp32362 2,704 MiB processmatches994x1482
NVIDIA L4 CUDA, torch 2.14.1+cu130, 2026-11-07Portablefp1675.2 987 MiB/1,132 MiB–994x1482
Kernelsfp1644.0 805 MiB/868 MiB–994x1482
NVIDIA L40S CUDA, torch 2.14.1+cu130, 2026-11-07Portablefp1624.9 987 MiB/1,132 MiB–994x1482
Kernelsfp1618.0 805 MiB/868 MiB–994x1482
NVIDIA A10 CUDA, torch 2.14.1+cu130, 2026-11-07Portablefp1656.7 987 MiB/1,132 MiB–994x1482
Kernelsfp1635.6 805 MiB/868 MiB–994x1482
NVIDIA A100-SXM4-40GB CUDA, torch 2.14.1+cu130, 2026-11-08Portablefp1626.5 987 MiB/1,132 MiB–994x1482
Kernelsfp1622.1 805 MiB/868 MiB–994x1482
NVIDIA RTX PRO 6000 CUDA, torch 2.14.1+cu130, 2026-11-08Portablefp1615.4 1,011 MiB/1,132 MiB–994x1482
Kernelsfp1611.0 829 MiB/908 MiB–994x1482

CUDA memory is torch’s peak allocated / reserved; Apple has no such counter, so the Mac rows show the process footprint (weights, Python and torch included). “Output vs reference” compares the device’s output with the stored golden output under the card’s tolerance; – = not checked on that device. Kernels “not implemented” = no kernels for this model on that backend yet: it runs portable there.

Licence and weights

Commercial use allowed

Code
MIT LICENSE
Weights
MIT source
Training data
Mixture of public synthetic and real RGB-D / multi-view datasets per the MoGe-2 and MoGe-3 papers

Our reading of the licences, not legal advice.

Weights

moge-2-vits-normal/model.pt 141 MB· hosted on Hugging Face

fastforks downloads and checks these files on first use (fastforks download moge-2-vits-normal).

Accuracy

The test set through the authors’ code and through ours: same checkpoint, same protocol. The distance between the original’s fp32 and fp16 runs is what half precision alone costs; our runs should stay within it.

NYUv2 test

AbsRel / delta1

RundtypeAbsRel ↓δ1 ↑Δ AbsRel vs fp32 originalimagesms / image
Original fp32fp320.04840.9757 reference 654 163
Original fp16fp160.04840.9757 ±0.0000 654 82.5
Portable fp32fp320.04840.9757 ±0.0000 654 137
Portablefp160.04840.9757 ±0.0000 654 63.5
Kernelsfp160.04840.9757 ±0.0000 654 31.5

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-06; portable fp32: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 6d3bab64, 2026-11-17; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.

iBims-1

AbsRel / delta1

RundtypeAbsRel ↓δ1 ↑Δ AbsRel vs fp32 originalimagesms / image
Original fp32fp320.04300.9793 reference 100 161
Original fp16fp160.04300.9793 ±0.0000 100 81.3
Portable fp32fp320.04300.9793 ±0.0000 100 136
Portablefp160.04300.9793 ±0.0000 100 62.8
Kernelsfp160.04300.9793 ±0.0000 100 32.7

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-06; portable fp32: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 6d3bab64, 2026-11-17; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.

ETH3D two-view train left

AbsRel / delta1training data not stated: possibly seen

RundtypeAbsRel ↓δ1 ↑Δ AbsRel vs fp32 originalimagesms / image
Original fp32fp320.07000.9412 reference 27 162
Original fp16fp160.07000.9411 ±0.0000 27 82.3
Portable fp32not measured
Portablefp160.07000.9411 ±0.0000 27 63.9
Kernelsfp160.07010.9412 −0.0001 27 71.0

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-06; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.

Middlebury 2014 trainingH left

AbsRel / delta1training data not stated: possibly seen

RundtypeAbsRel ↓δ1 ↑Δ AbsRel vs fp32 originalimagesms / image
Original fp32fp320.03560.9947 reference 15 163
Original fp16fp160.03560.9947 ±0.0000 15 83.5
Portable fp32not measured
Portablefp160.03560.9947 ±0.0000 15 64.0
Kernelsfp160.03560.9947 ±0.0000 15 102

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Original = the authors’ code, same checkpoint and protocol. Measured on original fp32, original fp16: RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-06; portable, kernels: RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.

NYUv2 test normals

mean / median angle deg, % < 11.25 / < 30 deg

Rundtypeang_mean ↑ang_median ↑a11.25 ↑a30 ↑imagesms / image
Original fp32not measured
Original fp16not measured
Portable fp32fp3215.879.2558.3283.97 654 164
Portablefp1615.879.2558.3283.97 654 80.7
Kernelsfp1615.879.2558.3283.97 654 40.3

The original code was not run on this dataset, so the table cannot show the distance to it.

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Normals: angle to the ground-truth normal over the pixels with ground truth (Marigold's evaluation sets), mean over images. Measured on RTX 5060 Ti, torch 2.14.1+cu130, fastforks 4e890760, 2026-11-12.

iBims-1 normals

mean / median angle deg, % < 11.25 / < 30 deg

Rundtypeang_mean ↑ang_median ↑a11.25 ↑a30 ↑imagesms / image
Original fp32not measured
Original fp16not measured
Portable fp32fp3215.957.1966.5683.74 100 163
Portablefp1615.957.1966.5683.74 100 79.5
Kernelsfp1615.957.1966.5683.74 100 39.3

The original code was not run on this dataset, so the table cannot show the distance to it.

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Measured on RTX 5060 Ti, torch 2.14.1+cu130, fastforks 4e890760, 2026-11-12.

Closeness to the original

Beyond the test-set scores: the raw output of ours against the original’s fp32 output on the same input, pixel by pixel. CPU, float32: max |diff| <= 0.005 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 1.5 % and p99 <= 1.5 % of the reference's median |value| (absolute error of the output depth).

CheckToleranceMeasured
CPU fp32, max |diff|≤ 0.005 (relative)not recorded
GPU default dtypes, mean |diff|≤ 0.015 (relative)0.00116 (relative)
GPU default dtypes, p99 |diff|≤ 0.015 (relative)0.0013 (relative)

Recorded for moge-2-vits-normal on RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-07, with these kernels: dpt_fusedfastbbsparse_conv

Details

MoGe-2 / MoGe-3 (Wang et al., Microsoft): monocular metric geometry; a DINOv2 ViT with conv heads predicts an affine point map, a validity mask, normals and a metric scale, and the focal length and shift are recovered from the points (MoGe-3 adds a sparse 3-D refiner).

Input / output
Input: one RGB image. Output: a depth or inverse-depth map.
What the numbers mean, and which models need the camera focal length

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).

Input size
any H x W >= 8, resized or padded inside the model (ViT patch 14); tested up to 3000x4000 (1,257 MiB allocated)
Options
  • resolution_level=90 to 9· changes the output
  • normals=false· changes the output
Outputs
depth, points, intrinsics, mask, normals
Checkpoint
arch="vits", version=2
CUDA kernels
  • fastbbfast backbone kernels: flash attention, window attention, instance norm and rotary embeddings for ViT and similar backbones
  • dpt_fusedfused fp16 matrix multiplies and convolutions for DINOv2 backbones and DPT depth heads
  • sparse_convMoGe-3: sparse 3D convolution
Apple (MPS)
sparse_metalMetal: sparse 3D convolution. fp32, fp16

Limitations

  • MoGe-3 compiles the dense network and leaves sparse voxel refinement eager. Its 100-image iBims-1 comparison used the default CUDA float16 mode.
  • At 994 x 1482, the portable and compiled MoGe-3 timings use the same synthetic image. The original timings use the Middlebury Motorcycle image.
Full reference, generated from the repository

Microsoft's MoGe: depth, 3D points, normals and focal length from one image.

Commercial use: yes.

variantinputbackendtime (ms)peak memory (MiB, allocated / reserved)
moge-2-vits-normalMiddlebury 2014 half (994x1482)original (fp32)
original
portable
kernels
portable + torch.compile
TensorRT fp16
162
82.6
61.6
33.3
43.9
24.5
1491 / 1628
1193 / 1268
1007 / 1172
932 / 1048
836 / 992
768 / 852
moge-2-vitb-normalsynthetic (994x1482)original (fp32)
original
portable
kernels
portable + torch.compile
TensorRT fp16
237
102
83.9
40.7
60.9
35.5
1773 / 1876
1469 / 1596
1163 / 1312
1093 / 1208
972 / 1172
1044 / 1092
moge-2-vitl-normalMiddlebury 2014 half (994x1482)original (fp32)
original
portable
kernels
portable + torch.compile
TensorRT fp16
501
171
144
73.4
124
75.6
2646 / 2818
2339 / 2438
1661 / 1836
1498 / 1610
1411 / 1634
1882 / 1936
moge-3-vitlMiddlebury 2014 half (994x1482)original (fp32)
original
portable
kernels
portable + torch.compile
TensorRT fp16
843
486
334
149
311
planned for 0.2.1
2812 / 2990
2505 / 2630
1678 / 1826
1626 / 1762
1500 / 1724
–
Details
  • Original code: https://github.com/microsoft/MoGe at 74fbce0
  • Code licence: MIT (permissive); files: LICENSE
  • Weights licence: MIT
  • Licence evidence: https://github.com/microsoft/MoGe/blob/main/README.md (License section) and the Hugging Face model cards (license: mit)
  • Training data: Mixture of public synthetic and real RGB-D / multi-view datasets per the MoGe-2 and MoGe-3 papers
  • Options: resolution_level = 9 (0 to 9), normals = False
  • Kernel libraries: fastbb, dpt_fused, sparse_conv
  • Variants: moge-2-vits-normal, moge-2-vitb-normal, moge-2-vitl-normal, moge-3-vitl
  • Not in this release (planned for 0.2.1): moge-3-vitl ONNX / TensorRT export: torch.export fails (TorchExportError, step 1/3 of the ONNX export)
  • Outputs: depth, points, intrinsics, mask, normals; kind: metric
  • Input size:
    • moge-2-vits-normal, moge-2-vitl-normal, moge-3-vitl: any H x W >= 8, resized or padded inside the model (ViT patch 14)
    • moge-2-vitb-normal: any H x W >= 1, resized or padded inside the model (ViT patch 14)
  • Tested up to (portable, fp16, RTX 5060 Ti 16 GB): moge-2-vits-normal 3000x4000 (1.2 GiB peak allocated, 1.4 reserved); moge-2-vitb-normal 3000x4000 (1.3 GiB peak allocated, 1.4 reserved); moge-2-vitl-normal 3000x4000 (2.4 GiB peak allocated, 2.5 reserved); moge-3-vitl 3000x4000 (2.6 GiB peak allocated, 3.0 reserved)
  • moge-2-vits-normal 994x1482, first call: portable + torch.compile 124 s (cold Inductor cache), TensorRT fp16 131 s (ONNX + engine).
  • moge-2-vitb-normal 994x1482, first call: portable + torch.compile 133 s (cold Inductor cache), TensorRT fp16 127 s (ONNX + engine).
  • moge-2-vitl-normal 994x1482, first call: portable + torch.compile 152 s (cold Inductor cache), TensorRT fp16 173 s (ONNX + engine).
  • moge-3-vitl 994x1482, first call: portable + torch.compile 136 s (cold Inductor cache).
  • moge-2-vits-normal load / first call: original 0.88 s / 1.501 s, portable 1.18 s / 0.298 s (518x770), kernels 0.21 s / 0.438 s (518x770)
  • moge-2-vitb-normal load / first call: original 1.26 s / 0.239 s, portable 0.49 s / 0.397 s (518x770), kernels 1.24 s / 0.285 s (518x770)
  • moge-2-vitl-normal load / first call: original 7.24 s / 0.171 s, portable 1.37 s / 0.144 s, kernels 0.63† s / 0.131 s
  • moge-3-vitl load / first call: original 10.6 s / 2.918 s, portable 2.31 s / 0.69 s (518x770), kernels 1.16 s / 0.742 s (518x770)
  • † measured in the speed sweep after the portable load (or an earlier size) in the same process: file cache and kernels are warm, so the time is lower than a fresh start.
  • Protocol, moge-2-vits-normal, moge-2-vitb-normal, moge-3-vitl: first call, 3 warm-up, peak reset, median of 3 rounds x median of 5
  • Protocol, moge-2-vitl-normal: median of 5 after 3 warm-up, uint8 CUDA input (synthetic for bench_inputs sizes), card defaults; gc before each load; peak = max over the first call and the timed calls

Measured

AbsRel / delta1 after a least-squares scale + shift fit (in disparity space for inverse-depth models), mean over images. Normals: angle to the ground-truth normal over the pixels with ground truth (Marigold's evaluation sets), mean over images. Original = the original code named under Details, same checkpoint and protocol.

variantdatasetoriginal fp32original fp16portableportable fp32kernelsportable + torch.compilenote
moge-2-vits-normalNYUv2 test, AbsRel / delta10.0484 / 0.97570.0484 / 0.97570.0484 / 0.97570.0484 / 0.97570.0484 / 0.9757
moge-2-vits-normaliBims-1, AbsRel / delta10.0430 / 0.97930.0430 / 0.97930.0430 / 0.97930.0430 / 0.97930.0430 / 0.9793
moge-2-vits-normalETH3D two-view train left, AbsRel / delta10.0700 / 0.94120.0700 / 0.94110.0700 / 0.94110.0701 / 0.9412possibly in the training data
moge-2-vits-normalMiddlebury 2014 trainingH left, AbsRel / delta10.0356 / 0.99470.0356 / 0.99470.0356 / 0.99470.0356 / 0.9947possibly in the training data
moge-2-vits-normalNYUv2 test normals, mean / median angle deg, % < 11.25 / < 30 deg15.87 / 9.25 / 58.32 / 83.9715.87 / 9.25 / 58.32 / 83.9715.87 / 9.25 / 58.32 / 83.97
moge-2-vits-normaliBims-1 normals, mean / median angle deg, % < 11.25 / < 30 deg15.95 / 7.19 / 66.56 / 83.7415.95 / 7.19 / 66.56 / 83.7415.95 / 7.19 / 66.56 / 83.74
moge-2-vitb-normalNYUv2 test, AbsRel / delta10.0404 / 0.98120.0404 / 0.98120.0404 / 0.98120.0404 / 0.98120.0404 / 0.9812
moge-2-vitb-normaliBims-1, AbsRel / delta10.0337 / 0.98550.0337 / 0.98540.0337 / 0.98540.0337 / 0.98550.0337 / 0.9854
moge-2-vitb-normalETH3D two-view train left, AbsRel / delta10.0667 / 0.94920.0667 / 0.94920.0667 / 0.94920.0663 / 0.9496possibly in the training data
moge-2-vitb-normalMiddlebury 2014 trainingH left, AbsRel / delta10.0260 / 0.99620.0260 / 0.99620.0260 / 0.99620.0261 / 0.9962possibly in the training data
moge-2-vitb-normalNYUv2 test normals, mean / median angle deg, % < 11.25 / < 30 deg15.25 / 8.61 / 60.32 / 84.9015.25 / 8.61 / 60.32 / 84.9015.25 / 8.61 / 60.33 / 84.90
moge-2-vitb-normaliBims-1 normals, mean / median angle deg, % < 11.25 / < 30 deg15.28 / 6.98 / 68.57 / 85.0115.28 / 6.97 / 68.57 / 85.0115.28 / 6.98 / 68.56 / 85.01
moge-2-vitl-normalNYUv2 test, AbsRel / delta10.0357 / 0.98360.0357 / 0.98360.0357 / 0.98360.0357 / 0.98360.0357 / 0.9836
moge-2-vitl-normaliBims-1, AbsRel / delta10.0261 / 0.98920.0261 / 0.98920.0261 / 0.98920.0261 / 0.98920.0261 / 0.9892
moge-2-vitl-normalETH3D two-view train left, AbsRel / delta10.0518 / 0.97890.0518 / 0.97890.0519 / 0.97890.0518 / 0.9790possibly in the training data
moge-2-vitl-normalMiddlebury 2014 trainingH left, AbsRel / delta10.0235 / 0.99720.0235 / 0.99720.0235 / 0.99720.0235 / 0.9972possibly in the training data
moge-2-vitl-normalNYUv2 test normals, mean / median angle deg, % < 11.25 / < 30 deg14.81 / 8.07 / 61.96 / 85.5414.81 / 8.07 / 61.96 / 85.5414.81 / 8.07 / 61.96 / 85.55
moge-2-vitl-normaliBims-1 normals, mean / median angle deg, % < 11.25 / < 30 deg14.70 / 6.49 / 70.49 / 85.8014.70 / 6.49 / 70.49 / 85.8014.69 / 6.49 / 70.49 / 85.80
moge-3-vitlNYUv2 test, AbsRel / delta10.0356 / 0.98360.0356 / 0.98360.0356 / 0.98360.0356 / 0.98360.0356 / 0.9836
moge-3-vitliBims-1, AbsRel / delta10.0261 / 0.98840.0261 / 0.98840.0261 / 0.98840.0261 / 0.98840.0261 / 0.98830.0261 / 0.9884
moge-3-vitlETH3D two-view train left, AbsRel / delta10.0496 / 0.98470.0496 / 0.98470.0496 / 0.98470.0496 / 0.9847possibly in the training data
moge-3-vitlMiddlebury 2014 trainingH left, AbsRel / delta10.0235 / 0.99700.0235 / 0.99700.0235 / 0.99700.0235 / 0.9970possibly in the training data

Measured on portable: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 24f9c9c8 / 3d00db1d / 4e890760 / 939adfe8, 2026-11-06 to 2026-11-12; portable32: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 24f9c9c8 / 4e890760 / 6d3bab64, 2026-11-12 to 2026-11-17; kernels: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 3d00db1d / 4e890760 / 939adfe8 / f3195fb2, 2026-11-06 to 2026-11-12; compile: NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks ed52b189, 2026-11-18.

Parity

CPU, float32: max |diff| <= 0.005 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 1.5 % and p99 <= 1.5 % of the reference's median |value| (absolute error of the output depth).

variantkernels error vs original, mean / p99GPU, torchkernels used
moge-2-vits-normal0.00116 / 0.0013RTX 5060 Ti, torch 2.14.1+cu130dpt_fused@48e0065bb7cf:sm_120, fastbb@7d058322634d:sm_120, sparse_conv@2feb2d5e0877:sm_120
moge-2-vitb-normal0.02172 / 0.02188RTX 5060 Ti, torch 2.14.1+cu130dpt_fused@48e0065bb7cf:sm_120, fastbb@7d058322634d:sm_120, sparse_conv@2feb2d5e0877:sm_120
moge-2-vitl-normal0.00382 / 0.00394RTX 5060 Ti, torch 2.14.1+cu130dpt_fused@48e0065bb7cf:sm_120, fastbb@7d058322634d:sm_120, sparse_conv@2feb2d5e0877:sm_120
moge-3-vitl0.00421 / 0.0049RTX 5060 Ti, torch 2.14.1+cu130dpt_fused@48e0065bb7cf:sm_120, fastbb@7d058322634d:sm_120, sparse_conv@2feb2d5e0877:sm_120

Limitations

  • MoGe-3 compiles the dense network and leaves sparse voxel refinement eager. Its 100-image iBims-1 comparison used the default CUDA float16 mode.
  • At 994 x 1482, the portable and compiled MoGe-3 timings use the same synthetic image. The original timings use the Middlebury Motorcycle image.

Example output

One fixed input through this checkpoint, the same for every model of the task (kernels · float16 · NVIDIA GeForce RTX 5060 Ti).

Input
Output
A cup of coffee, input
Input · 100%

Input: A cup of coffee, Rachel Michetti (CC0).