DEFOM-Stereo (fastforks)

DEFOM-Stereo (Jiang et al. 2025): Depth Anything V2 (ViT-L) features and initial disparity, then RAFT-style ConvGRU updates with scale updates, 32 iterations.

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/Insta360-Research-Team/DEFOM-Stereo at 5b27591d8395.

Task: stereo. Tier: stable. Outputs: disparity. Backends: portable, kernels.

Input: a rectified left / right image pair. Output: a disparity map (pixels, left view).

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
defom-stereo-vitl-sceneflow@1	defomstereo_vitl_sceneflow.pth	0be5c5ddd25e	CC-BY-NC-4.0	no

Licence

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

Training data: Freiburg SceneFlow (FlyingThings3D, Driving, Monkaa).

• Licence note: The weights embed Depth Anything V2 ViT-L weights released under CC-BY-NC-4.0, so they are non-commercial even though the code is MIT.

Licence evidence: https://github.com/Insta360-Research-Team/DEFOM-Stereo/blob/5b27591d83956e5d329ff66704271a9e68b1c75f/README.md

Input size

any H x W >= 32, padded to a multiple of 32 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): defom-stereo-vitl-sceneflow 1988x2880 (12.0 GiB peak allocated, 14.5 reserved)

Speed and memory

Stereo, 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
defom-stereo-vitl-sceneflow	4175.5	–	1530.89	658.71	–	2.3x	–	4362 / 4838	3154 / 3566

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
defom-stereo-vitl-sceneflow	9.54 (375x1242)	2.31	1.18†	1.04 (375x1242)	1.53†	0.66†

Fixed-size speed and peak memory, all models

model	mode	640x480	1280x720	1920x1080	12 MP 4000x3000
defom-stereo-vitl-sceneflow	portable	263.9 ms / 2070 / 2368	860.4 ms / 3248 / 4044	2169.2 ms (n=3) / 5378 / 7082	OOM
defom-stereo-vitl-sceneflow	kernels	120.5 ms / 1357 / 1702	376.5 ms / 2275 / 3190	936.5 ms / 2991 / 3578	5950.0 ms (n=3) / 11923 / 15526

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

On public datasets

Original = the authors' code, same checkpoint and protocol.

variant	dataset	original	portable	kernels	note
defom-stereo-vitl-sceneflow	ETH3D two-view train, bad > 1 px % / EPE	2.40 / 0.337	2.42 / 0.362	2.44 / 0.341
defom-stereo-vitl-sceneflow	Middlebury 2014 trainingH, bad > 2 px % / EPE	6.03 / 0.837	6.14 / 0.848	6.02 / 0.834

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, 2026-10-27.

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, portable bfloat16 and kernels float16 (the defaults): mean |diff| <= 0.02 px and p99 <= 0.25 px.

Limitations

• Only the ViT-L SceneFlow checkpoint is included (ViT-S / ViT-B and the Middlebury / ETH3D / KITTI fine-tunes are not).
• Portable is faster than the original code but allocates more memory (about 1.25x at 994x1482); kernels is several times faster and needs a fraction of the memory.

Usage

from fastforks import stereo

model = stereo.load("defom-stereo-vitl-sceneflow")  # portable PyTorch on the GPU if there is one, else the CPU
result = model("left.png", "right.png")
result.disparity  # [H, W] float32 tensor, input resolution
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: defom-stereo-vitl-sceneflow.

Kernels backend

stereo.load("defom-stereo-vitl-sceneflow", mode="kernels") runs our CUDA kernels (igev, cutlass_conv, cudnn_hacc, fastbb). 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.