CFNet (fastforks)

CFNet (Shen et al. 2021): cascade and fused cost volumes (fused 1/8-1/32 3-D volumes), then two refinement stages that adapt to the uncertainty of the disparity (max disparity 192).

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/gallenszl/CFNet at 982230dd3605.

Task: stereo. Tier: experimental. 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
cfnet-sceneflow@1	sceneflow_pretraining.ckpt	07ea542501a7	MIT (assumed from code)	yes

Licence

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

Code licence: MIT; the authors do not state a separate weights licence.

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

Licence evidence: https://github.com/gallenszl/CFNet/blob/982230dd360583fc35743d0b7ed528982fe5fd38/README.md#L72

Input size

any H x W >= 64, padded to a multiple of 64 Tested up to (portable, fp16, RTX 5060 Ti 16 GB): cfnet-sceneflow 1988x2880 (8.2 GiB peak allocated, 11.5 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
cfnet-sceneflow	660	–	484	401	–	1.2x	–	3441 / 4376	3662 / 4718

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
cfnet-sceneflow	0.39 (375x1242)	0.94	0.21†	0.24 (375x1242)	0.48†	0.40†

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)
cfnet-sceneflow	994x1482	484	249.2	53.0	170.9	139	401	3441 / 4376	1026 / 1398	1238

Fixed-size speed and peak memory, all models

model	mode	480x640	720x1280	1080x1920	12 MP 3000x4000
cfnet-sceneflow	portable	73.6 ms / 1070 / 1588	284.0 ms / 2640 / 4342	758.8 ms / 4508 / 5956	OOM
cfnet-sceneflow	kernels	59.3 ms / 1118 / 1618	229.9 ms / 2778 / 4362	634.1 ms / 4801 / 5976	OOM

Apple silicon (MPS)

model	portable	kernels	speed-up
cfnet-sceneflow	2228 ms / 5463 MiB (float16)	758 ms / 2931 MiB	2.94x

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 fp32	original fp16	portable	kernels	note
cfnet-sceneflow	ETH3D two-view train, bad > 1 px % / EPE	5.59 / 0.417	99.92 / 13.132 (NaN / inf output)	5.59 / 0.417	5.60 / 0.416
cfnet-sceneflow	Middlebury 2014 trainingH, bad > 2 px % / EPE	21.31 / 6.445	100.00 / 71.272 (NaN / inf output)	21.32 / 6.447	21.32 / 6.444

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, 2026-11-05.

Parity with the original code

Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.05 px and p99 <= 0.5 px.

Usage

from fastforks import stereo

model = stereo.load("cfnet-sceneflow")  # mode="auto": the preferred backend from recorded measurements
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: cfnet-sceneflow.

Kernels backend

stereo.load("cfnet-sceneflow", mode="kernels") runs our CUDA kernels (igev, cutlass_conv, 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.