DepthSplat (fastforks)
DepthSplat (Xu et al. 2025): feed-forward 3D Gaussian splatting and depth from posed, calibrated images. A multi-view cost volume (UniMatch transformer, 2D U-Net) fused with a monocular DINOv2 ViT-S predicts every view's depth; a DPT feature upsampler and a Gaussian head predict one Gaussian per pixel (opacity, scale, rotation, degree-2 SH) in the given world frame (result.gaussians, .save("scene.ply")). Forward pass only: no renderer.
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/cvg/depthsplat at 2dad25a.
Author weight repos: https://huggingface.co/haofeixu/depthsplat.
Task: multiview. Tier: experimental. Outputs: depth, confidence, gaussians. Backends: portable, kernels.
Input: several images of one scene (posed and calibrated for multi-view stereo, unposed otherwise). Output: depth maps and, where the model predicts them, cameras and 3-D points.
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 |
|---|---|---|---|---|
depthsplat-small-re10k@1 |
depthsplat-gs-small-re10k-256x256-view2-cfeab6b1.pth | cfeab6b10077 |
MIT | yes |
Licence
Original-code strings, recorded as they are (research-use package; check them before any other use). Code: MIT. Licence files: LICENSE.
Training data: RealEstate10K (camera trajectories of YouTube real-estate videos) (RealEstate10K (YouTube videos, camera poses from ORB-SLAM; research use per its release), from a UniMatch depth / Depth Anything V2 Small initialisation.).
- Licence note: Only the Small RealEstate10K checkpoint is shipped: MIT code and weights, its monocular branch initialised from Depth Anything V2 Small (Apache-2.0). The Base / Large checkpoints build on Depth Anything V2 Base / Large (CC-BY-NC-4.0) and are not offered.
Licence evidence: https://github.com/cvg/depthsplat/blob/2dad25a7b9ba7c5537aba08463ef88d8511a06be/LICENSE and https://github.com/cvg/depthsplat/blob/2dad25a7b9ba7c5537aba08463ef88d8511a06be/MODEL_ZOO.md and https://huggingface.co/haofeixu/depthsplat/tree/ed5116c11f7932fff3714989e2b65716730f6f87
Input size
any H x W >= 1, resized or padded inside the model
Tested up to (portable, fp16, RTX 5060 Ti 16 GB): depthsplat-small-re10k 3000x4000 (7.4 GiB peak allocated, 8.6 reserved)
Speed and memory
Multiview, 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 |
|---|---|---|---|---|---|---|---|---|---|---|
depthsplat-small-re10k |
336x518 | β | 126 | 119 | 112 | 1.1x | 1.1x | 5694 / 5892 | 5432 / 5606 | 5434 / 5606 |
Startup and first call
| model | load (s): original | portable | kernels | first call (s): original | portable | kernels |
|---|---|---|---|---|---|---|
depthsplat-small-re10k |
1.83 (336x518) | 0.22 (336x518) | 1.47 (336x518) | 2.97 (336x518) | 0.97 (336x518) | 0.71 (336x518) |
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) |
|---|---|---|---|---|---|---|---|---|---|---|---|
depthsplat-small-re10k |
336x518 | 119 ms | 82.2 ms | 86.4 ms | 112 ms | 80.0 s | 224 s | 5432 / 5606 | 5432 / 5526 | 5434 / 5606 | 3148 / 3194 |
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
640x512 crops, the 7 most co-visible views as sources; GT = depth rendered from the structured-light points. AbsRel / delta1 over GT pixels, after per-image median scaling unless the model takes the cameras (calibrated MVS: cameras, error in mm; PatchMatchNet also the depth range 425-935 mm, MVSAnywhere its own range from the camera baselines). Original = the original code named under Details, same checkpoint and protocol.
| variant | dataset | original fp32 | original fp16 | portable | kernels | note |
|---|---|---|---|---|---|---|
depthsplat-small-re10k |
DTU scans 1, 9, 24, 33, 75, view 24 + 7 sources, AbsRel / delta1 % / MAE mm | 0.0992 / 98.72 / 66.52 | 0.0974 / 98.69 / 65.37 | 0.0974 / 98.70 / 65.36 | 0.0974 / 98.69 / 65.34 |
Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 62af007c, 2026-11-12.
Parity with the original code
Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 1.6e-05 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.0035 and p99 <= 0.017 (absolute error of the output depth).
Usage
from fastforks import multiview
model = multiview.load("depthsplat-small-re10k") # mode="auto": the preferred backend from recorded measurements
result = model(["view0.jpg", "view1.jpg", "view2.jpg"]) # the views of one scene, no cameras needed
result.depth # [V, H, W] float32 tensor; result.kind says what the maps mean
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: depthsplat-small-re10k.
Kernels backend
multiview.load("depthsplat-small-re10k", mode="kernels") runs our CUDA kernels (fastbb, 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.