HITNet (fastforks)

HITNet (Tankovich et al. 2021): slanted tile hypotheses propagated and refined from coarse to fine; the authors' ETH3D (gray, max disparity 128) and Middlebury (half resolution, max disparity 400) graphs.

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/google-research/google-research at f200f2b89d29.

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
hitnet-eth3d@1	eth3d.pb	ab34a01de126	Apache-2.0 (assumed from code)	yes
hitnet-middlebury-d400@1	middlebury_d400.pb	cc386fbb3181	Apache-2.0 (assumed from code)	yes

Licence

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

Code licence: Apache-2.0; the authors do not state a separate weights licence.

Training data: not stated by the authors.

• Licence note: The authors state no weights licence; the checkpoints are fetched from the PINTO model-zoo mirror (a third-party host), not from Google, and the code licence is assumed to cover them.

Licence evidence: https://github.com/google-research/google-research/blob/f200f2b89d29484bf0dd0fef0c3979ac8bed9af4/LICENSE

Input size

any H x W >= 8, resized or padded inside the model Tested up to (portable, fp16, RTX 5060 Ti 16 GB): hitnet-eth3d 3000x4000 (6.4 GiB peak allocated, 10.8 reserved); hitnet-middlebury-d400 1988x2880 (6.4 GiB peak allocated, 10.1 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
hitnet-eth3d	–	–	67.78	23.16	–	2.9x	–	860 / 1090	647 / 1248
hitnet-middlebury-d400	–	–	176.0	52.73	–	3.3x	–	1744 / 2414	1263 / 2368

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
hitnet-eth3d	–	0.09	0.07†	–	0.08†	0.03†
hitnet-middlebury-d400	–	0.13	0.10†	–	0.18†	0.05†

Fixed-size speed and peak memory, all models

model	mode	640x480	1280x720	1920x1080	12 MP 4000x3000
hitnet-eth3d	portable	18.1 ms / 221 / 336	40.2 ms / 659 / 1004	92.6 ms / 1134 / 2438	579.4 ms / 6558 / 11110
hitnet-eth3d	kernels	5.2 ms / 140 / 470	14.5 ms / 392 / 482	30.8 ms / 836 / 1004	181.7 ms / 4814 / 5758
hitnet-middlebury-d400	portable	35.8 ms / 602 / 838	108.8 ms / 1067 / 2222	251.7 ms / 2392 / 3814	OOM
hitnet-middlebury-d400	kernels	11.7 ms / 266 / 966	33.1 ms / 759 / 978	74.2 ms / 1695 / 2248	426.2 ms / 9604 / 12664

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	portable	kernels	note
hitnet-eth3d	ETH3D two-view train, bad > 1 px % / EPE	0.93 / 0.164	0.93 / 0.164	trained on this dataset's public images; not a zero-shot result
hitnet-eth3d	Middlebury 2014 trainingH, bad > 2 px % / EPE	17.25 / 8.340	17.30 / 8.356
hitnet-middlebury-d400	ETH3D two-view train, bad > 1 px % / EPE	7.01 / 0.899	6.98 / 0.872
hitnet-middlebury-d400	Middlebury 2014 trainingH, bad > 2 px % / EPE	7.54 / 1.187	7.52 / 1.184	trained on this dataset's public images; not a zero-shot result

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| <= 0.02 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.15 px and p99 <= 1.0 px.

Limitations

• CUDA output (float16, kernels, float32 with PyTorch's default TF32 convolutions) differs from the CPU by ~0.2 px mean / 4 px p99 at 720x1280: the argmin flips on near-equal tile costs. Full float32 (torch.backends.cudnn.allow_tf32 = False) matches the CPU to 0.00002 px.
• The weights are the authors' frozen TensorFlow graphs read without TensorFlow; the FlyingThings and Middlebury d160 / d288 graphs are not included.
• The original code is TensorFlow and was not timed.

Usage

from fastforks import stereo

model = stereo.load("hitnet-eth3d")  # 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: hitnet-eth3d, hitnet-middlebury-d400.

Kernels backend

stereo.load("hitnet-eth3d", mode="kernels") runs our CUDA kernels (hitnet, hitnet_conv). 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.