SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2501.09466 · CVPR · 2025

DEFOM-Stereo: Depth Foundation Model Based Stereo Matching

Rui Xu, Rui Huang, Laiyan Ding, Zhiqiang Lou, Wenjie Jiang, Insta360 Research, Hualie Jiang, Minglang Tan

arXiv · PDF · Open in the Atlas

Code that ran

11 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 6 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2501.09466.json · badge: /api/ran/2501.09466.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
insta360-research-team/defom-stereo canonical (classifier label)
classifier basis not recorded on the link
11 ran · 2 no recorded run
FunctionStatusWhere it lives
StrToBytes Ran
ran on a synthesized input
insta360-research-team/defom-stereo/make_submission.py
code served (permissive licence) · get_code("b7e56726f74ed154")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bytes. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import sys

def StrToBytes(text):
    if sys.version_info[0] == 2:
        return text
    else:
        return bytes(text, 'UTF-8')

ARGS = {
    'text': 'Hello World'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bytes — a bytes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a bytes output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
count_parameters Ran
ran, honoured the contract we drafted
insta360-research-team/defom-stereo/make_submission.py
code served (permissive licence) · get_code("f6b944f50d3f15ae")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

ARGS = {
    "model": model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
count_parameters Ran
ran on a synthesized input
insta360-research-team/defom-stereo/evaluate_stereo.py
code served (permissive licence) · get_code("32cf80406c31514c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
        # Make one parameter non-trainable
        self.linear2.bias.requires_grad = False
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

ARGS = {
    "model": model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
forward_interpolate Ran
ran; our contract draft was wrong, not the code
insta360-research-team/defom-stereo/core/utils/utils.py
code served (permissive licence) · get_code("f7edf6ef2466762a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from scipy import interpolate

torch.manual_seed(0)

# Create a small flow tensor with shape (2, height, width)
# The function expects flow[0] to be dx and flow[1] to be dy
height = 4
width = 3
flow = torch.randn(2, height, width)

ARGS = {
    "flow": flow
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_danv2_io_size Ran
ran on a synthesized input
insta360-research-team/defom-stereo/core/utils/utils.py
code served (permissive licence) · get_code("4ad81b89a2bfe741")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    "h": 1024,
    "w": 768,
    "nds": 3,
    "max_i_size": 2688,
    "multiple_of": 14,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
interp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
insta360-research-team/defom-stereo/core/update.py
pointer only (licence: MIT); the stored copy get_code serves, from NVlabs/FoundationStereo, is pointer only (licence: NOASSERTION) · get_code("08bedbc0c0447f5d")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from NVlabs/FoundationStereo, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

pool2x Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
insta360-research-team/defom-stereo/core/update.py
pointer only (licence: MIT); the stored copy get_code serves, from NVlabs/FoundationStereo, is pointer only (licence: NOASSERTION) · get_code("25894ad2a73a8142")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from NVlabs/FoundationStereo, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

pool4x Ran
ran; our contract draft was wrong, not the code
insta360-research-team/defom-stereo/core/update.py
pointer only (licence: MIT); the stored copy get_code serves, from bartn8/stereoanywhere, is pointer only (licence: NOASSERTION) · get_code("771e8ac695af28a9")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from bartn8/stereoanywhere, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

readFlowKITTI Ran
ran; our contract draft was wrong, not the code
insta360-research-team/defom-stereo/core/utils/frame_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from VQE-CPGA/CPGA, is pointer only (licence: NONE) · get_code("04de40832d91cecb")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from VQE-CPGA/CPGA, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

readPFM Ran
ran, honoured the contract we drafted
insta360-research-team/defom-stereo/core/utils/frame_utils.py
code served (permissive licence) · get_code("7c3fbc1f552de802")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (3, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import tempfile
import os

# Create a minimal valid PFM file
# PFM format: header (PF or Pf), width height, scale, then binary float data

def create_test_pfm():
    """Create a minimal valid PFM file for testing"""
    with tempfile.NamedTemporaryFile(mode='wb', delete=False, suffix='.pfm') as f:
        temp_path = f.name
        
        # Write header for color image (PF)
        f.write(b'PF\n')
        
        # Write dimensions: width=4, height=3
        f.write(b'4 3\n')
        
        # Write scale (positive means big-endian)
        f.write(b'1.0\n')
        
        # Write binary float data: 4*3*3 = 36 floats for color image
        data = np.ones((3, 4, 3), dtype='>f')  # big-endian floats
        data.tofile(f)
    
    return temp_path

# Create the test file
test_file_path = create_test_pfm()

ARGS = {
    'file': test_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (3, 4, 3) — shape (3, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
transfer_color Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
insta360-research-team/defom-stereo/core/utils/augmentor.py
code served (permissive licence) · get_code("4524aae75dcbb73f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (4, 3, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from skimage import color

# Create a small test image in RGB format
# The function expects an RGB image that will be converted to LAB
# Using shape (height, width, channels) where channels=3 for RGB
np.random.seed(0)
image = np.random.rand(4, 3, 3).astype(np.float32)  # 4x3 spatial dims, 3 RGB channels

# Compute style statistics from a reference image
reference_image = np.random.rand(5, 6, 3).astype(np.float32)  # Different size reference
reference_image_lab = color.rgb2lab(reference_image)
style_mean = np.mean(reference_image_lab, axis=(0, 1), keepdims=True)
style_stddev = np.std(reference_image_lab, axis=(0, 1), keepdims=True)

ARGS = {
    "image": image,
    "style_mean": style_mean,
    "style_stddev": style_stddev,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, skimage — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (4, 3, 3) — shape (4, 3, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
bilinear_sampler Not yet run
unverified (harvested, no recorded run)
insta360-research-team/defom-stereo/core/utils/utils.py
code served (permissive licence) · get_code("c9398c239fc37bcb")
How it ran

no invocation recorded.

readFlow Not yet run
unverified (harvested, no recorded run)
insta360-research-team/defom-stereo/core/utils/frame_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from NVIDIA/flownet2-pytorch, is pointer only (licence: NOASSERTION) · get_code("1662ea0b62b0229c")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from NVIDIA/flownet2-pytorch, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 10 are also present in other papers whose code we hold (10 with at least one paper dated earlier, 2 the same month, 2 later) and 3 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
count_parameters 180 other papers
157 dated earlier, 21 later, 2 same year, order not determinable · 8 only as an assembly brick
1705.08415 (2017-05)
Supervised Community Detection with Line Graph Neural Networks
1707.01926 (2017-07)
Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting
1808.09367 (2018-08)
Deriving Machine Attention from Human Rationales
and 177 more not named here
pool2x 8 other papers
6 dated earlier, 1 the same month, 1 later · 1 only as an assembly brick
Zeng_Parameterized_Cost_Volume_for_Stereo_Matching_ICCV_2023_paper (2023 (year only))
Parameterized Cost Volume for Stereo Matching
2312.02155 (2023-12)
GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis
2403.00486 (2024-03)
Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching
and 5 more not named here
readFlow 8 other papers
8 dated earlier
1612.01925 (2016-12)
FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
2007.12622 (2020-07)
BMBC:Bilateral Motion Estimation with Bilateral Cost Volume for Video Interpolation
2010.14851 (2020-10)
Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation
and 5 more not named here
interp 7 other papers
6 dated earlier, 1 the same month
Zeng_Parameterized_Cost_Volume_for_Stereo_Matching_ICCV_2023_paper (2023 (year only))
Parameterized Cost Volume for Stereo Matching
2312.02155 (2023-12)
GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis
2403.00486 (2024-03)
Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching
and 4 more not named here
pool4x 6 other papers
6 dated earlier
Zeng_Parameterized_Cost_Volume_for_Stereo_Matching_ICCV_2023_paper (2023 (year only))
Parameterized Cost Volume for Stereo Matching
2312.02155 (2023-12)
GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis
2403.00486 (2024-03)
Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching
and 3 more not named here
readPFM 6 other papers
6 dated earlier
2010.14851 (2020-10)
Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation
2104.13918 (2021-04)
High-Resolution Optical Flow from 1D Attention and Correlation
2109.15321 (2021-09)
Sensor-Guided Optical Flow
and 3 more not named here
readFlowKITTI 5 other papers
5 dated earlier
2010.14851 (2020-10)
Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation
2104.13918 (2021-04)
High-Resolution Optical Flow from 1D Attention and Correlation
2109.15321 (2021-09)
Sensor-Guided Optical Flow
and 2 more not named here
forward_interpolate 4 other papers
3 dated earlier, 1 same year, order not determinable
2109.15321 (2021-09)
Sensor-Guided Optical Flow
2308.13133 (2023-08)
AccFlow: Backward Accumulation for Long-Range Optical Flow
and 1 more not named here
bilinear_sampler 1 other paper
1 dated earlier
Zeng_Parameterized_Cost_Volume_for_Stereo_Matching_ICCV_2023_paper (2023 (year only))
Parameterized Cost Volume for Stereo Matching
transfer_color 1 other paper
1 dated earlier
Zeng_Parameterized_Cost_Volume_for_Stereo_Matching_ICCV_2023_paper (2023 (year only))
Parameterized Cost Volume for Stereo Matching

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Stereo matching is a key technique for metric depth estimation in computer vision and robotics. Real-world challenges like occlusion and non-texture hinder accurate disparity estimation from binocular matching cues. Recently, monocular relative depth estimation has shown remarkable generalization using vision foundation models. Thus, to facilitate robust stereo matching with monocular depth cues, we incorporate a robust monocular relative depth model into the recurrent stereo-matching framework, building a new framework for depth foundation model-based stereomatching, DEFOM-Stereo. In the feature extraction stage, we construct the combined context and matching feature encoder by integrating features from conventional CNNs and DEFOM. In the update stage, we use the depth predicted by DEFOM to initialize the recurrent disparity and introduce a scale update module to refine the disparity at the correct scale. DEFOM-Stereo is verified to have much stronger zero-shot generalization compared with SOTA methods. Moreover, DEFOM-Stereo achieves top performance on the KITTI 2012, KITTI 2015, Middlebury, and ETH3D benchmarks, ranking 1 st on many metrics. In the joint evaluation under the robust vision challenge, our model simultaneously outperforms previous models on the individual benchmarks, further demonstrating its outstanding capabilities.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2501.09466")
get_code_for_paper("2501.09466")
have("2501.09466")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2501.09466.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2501.09466.svg)](https://syntology.ai/paper/2501.09466)

Connect an agent — have() is free.