SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2209.07923 · ECCV · 2022

A Deep Moving-camera Background Model

Ron Shapira, Oren Freifeld, Guy Erez

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2209.07923.json · badge: /api/ran/2209.07923.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
BGU-CS-VIL/DeepMCBM community
Papers with Code archive
8 ran · 2 no recorded run
FunctionStatusWhere it lives
affine_warp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("75786188ad2fbe6e")
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. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# Define the function that will be tested
device = torch.device('cpu')

def get_grids(shape, homogenous=False, device='cpu') -> torch.Tensor:
    N, C, H, W = shape
    y, x = torch.meshgrid([torch.linspace(-1, 1, H), torch.linspace(-1, 1, W)])
    x, y = (x.flatten(), y.flatten())
    if homogenous:
        grid = torch.stack([x, y, torch.ones(x.shape[0])], dim=0)
    else:
        grid = torch.stack([x, y], dim=0)
    grid = grid.repeat(N, 1, 1)
    return grid.to(device)

def affine_warp(theta, shape, exp=True, grid=None, return_homogenous_grid=False, inverse=False, global_transform=None, device='cpu') -> torch.Tensor:
    """
    Args:
        theta: (N, 6)
        shape: (N, C, H, W)
        grid: (N, 2, H*W)
        inverse: bool
        exp: use matrix exponental - when learning Lie Algebra via STN (bool)
    Returns:
        grid: transofrmed grid (N, 2, H*W)
        use grid.permute(0,2,1).view(-1, H, W, 2) to get back to image coordinates

    """
    N, C, H, W = shape
    if grid is None:
        grid = get_grids(shape, homogenous=True, device=device)
    elif grid.shape[1] == 2:
        x, y = (grid[:, 0, :].reshape(N, -1), grid[:, 1, :].reshape(N, -1))
        x, y = (x.reshape(N, -1), y.reshape(N, -1))
        grid = torch.stack([x, y, torch.ones((N, x.shape[-1]), device=device)], dim=1)
    theta = theta.reshape(-1, 2, 3)
    theta = F.pad(theta, (0, 0, 0, 1))
    if exp:
        theta = torch.matrix_exp(theta)
    else:
        theta[:, -1, -1] = 1
    if global_transform is not None:
        theta = theta @ global_transform
    if inverse:
        theta = torch.inverse(theta)
    grid_t = theta.matmul(grid)
    if not return_homogenous_grid:
        grid_t = grid_t[:, :2, :]
    return (theta, grid_t)

# Create valid arguments
N, C, H, W = 2, 3, 4, 5
theta = torch.randn(N, 6)
shape = (N, C, H, W)
grid_input = None
return_homogenous_grid = False
inverse = False
global_transform = torch.eye(3).unsqueeze(0).repeat(N, 1, 1)
device_arg = 'cpu'
exp = True

ARGS = {
    'theta': theta,
    'shape': shape,
    'exp': exp,
    'grid': grid_input,
    'return_homogenous_grid': return_homogenous_grid,
    'inverse': inverse,
    'global_transform': global_transform,
    'device': device_arg,
}
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 — a tuple output has no array shape of its own; none is 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
create_pixel_stack Ran
ran; our contract draft was wrong, not the code
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("743452a3eb2ee8c1")
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. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal STN (Spatial Transformer Network) mock
class MockSTN(nn.Module):
    def __init__(self):
        super().__init__()
    
    def forward(self, image):
        # The function expects stn(image) to return (image_out, mask_out, transform)
        batch_size = image.shape[0]
        # Return transformed image, mask, and transform matrix
        image_out = image * 0.9  # Simple transformation
        mask_out = torch.ones(batch_size, 1, 4, 4)
        transform = torch.eye(3).unsqueeze(0).expand(batch_size, -1, -1)
        return image_out, mask_out, transform

# Create a minimal data loader (list of batches)
data_loader = [
    torch.randn(2, 3, 4, 4),  # batch of 2 images with 3 channels, 4x4 spatial dims
    torch.randn(2, 3, 4, 4),  # another batch
]

# Create STN instance
stn = MockSTN()

ARGS = {
    "data_loader": data_loader,
    "stn": stn,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
get_grids Ran
ran; our contract draft was wrong, not the code
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("f2974723c2e7d212")
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, 2, 20) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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

torch.manual_seed(0)

ARGS = {
    'shape': (2, 3, 4, 5),
    'homogenous': False,
    'device': 'cpu'
}
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 Tensor, ndim 3, dtype family float, shape (2, 2, 20) — shape (2, 2, 20) 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
homography_warp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("ad403c0a82e10079")
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. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch

torch.manual_seed(0)

# theta: (N, 8) - 8 parameters for perspective transformation
theta = torch.randn(2, 8)

# shape: (N, C, H, W)
shape = (2, 3, 4, 5)

# exp: bool - use matrix exponential
exp = True

# grid: (N, 2, H*W) or None - pre-computed grid
grid = None

# return_homogenous_grid: bool
return_homogenous_grid = False

# inverse: bool - whether to invert the transformation matrix
inverse = False

# device: str
device = 'cpu'

ARGS = {
    'theta': theta,
    'shape': shape,
    'exp': exp,
    'grid': grid,
    'return_homogenous_grid': return_homogenous_grid,
    'inverse': inverse,
    'device': device,
}
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 — a tuple output has no array shape of its own; none is 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
trim_pixel_stack Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("55a9adb3d42575f9")
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 5, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 numpy as np

torch.manual_seed(0)

# Create small test tensors with shape (batch, channels, height, width)
# Using distinct sizes: 2, 3, 4, 5
batch_size = 2
channels = 3
height = 4
width = 5

# image_stack: shape (batch, channels, height, width)
image_stack = torch.randn(batch_size, channels, height, width)

# mask_stack: shape (batch, channels, height, width) with values in [0, 1]
mask_stack = torch.rand(batch_size, channels, height, width)

# trim_percentage: float between 0 and 1
trim_percentage = 0.1

ARGS = {
    'image_stack': image_stack,
    'mask_stack': mask_stack,
    'trim_percentage': trim_percentage,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 5, dtype family float — 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
vec_to_perpective_matrix Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("57a0fe3493fb5ce4")
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, 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 torch

torch.manual_seed(0)

# Create a small test tensor with shape (batch_size=2, features=8)
# The function expects vec to have shape (N, 8) and will:
# 1. Concatenate ones to make (N, 9)
# 2. Reshape to (N, 9)
# 3. Reshape to (N, 3, 3)
# So we need vec.shape[0] * 9 = vec.shape[0] * 3 * 3, which checks out
vec = torch.randn(2, 8, dtype=torch.float32)

ARGS = {
    "vec": vec
}
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 Tensor, ndim 3, dtype family float, shape (2, 3, 3) — shape (2, 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
warp_inv Ran
ran; our contract draft was wrong, not the code
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("65eb6ff441527c87")
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 4, dtype family float, shape (2, 3, 4, 5) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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.functional as F

torch.manual_seed(0)

# Helper function from the code
def get_grids(shape, homogenous=False, device='cpu') -> torch.Tensor:
    N, C, H, W = shape
    y, x = torch.meshgrid([torch.linspace(-1, 1, H), torch.linspace(-1, 1, W)], indexing='ij')
    x, y = (x.flatten(), y.flatten())
    if homogenous:
        grid = torch.stack([x, y, torch.ones(x.shape[0])], dim=0)
    else:
        grid = torch.stack([x, y], dim=0)
    grid = grid.repeat(N, 1, 1)
    return grid.to(device)

# Helper function from the code
def vec_to_perpective_matrix(vec):
    out = torch.cat((vec, torch.ones((vec.shape[0], 1), dtype=vec.dtype, device=vec.device)), dim=1).reshape(vec.shape[0], -1)
    return out.reshape(-1, 3, 3)

# Helper function from the code
def homography_warp(theta, shape, exp=True, grid=None, return_homogenous_grid=False, inverse=False, device='cpu') -> torch.Tensor:
    assert len(shape) == 4, 'Shape must be 4D: N, C, H, W'
    N, C, H, W = shape
    theta = vec_to_perpective_matrix(theta)
    if exp:
        theta[:, 2, 2] = -theta[:, 0, 0] - theta[:, 1, 1]
        theta = torch.matrix_exp(theta)
    if inverse:
        theta = torch.inverse(theta)
    if grid is None:
        grid = get_grids(shape, homogenous=True, device=device)
    elif grid.shape[1] == 2:
        x, y = (grid[:, 0, :].reshape(N, -1), grid[:, 1, :].reshape(N, -1))
        x, y = (x.reshape(N, -1), y.reshape(N, -1))
        grid = torch.stack([x, y, torch.ones((N, x.shape[-1]), device=device)], dim=1)
    grid_t = theta.matmul(grid)
    grid_t = grid_t[:, :2, :] / (grid_t[:, 2, :].unsqueeze(1) + 1e-09)
    return (theta, grid_t)

# Helper function from the code
def affine_warp(theta, shape, exp=True, grid=None, return_homogenous_grid=False, inverse=False, global_transform=None, device='cpu') -> torch.Tensor:
    N, C, H, W = shape
    if grid is None:
        grid = get_grids(shape, homogenous=True, device=device)
    elif grid.shape[1] == 2:
        x, y = (grid[:, 0, :].reshape(N, -1), grid[:, 1, :].reshape(N, -1))
        x, y = (x.reshape(N, -1), y.reshape(N, -1))
        grid = torch.stack([x, y, torch.ones((N, x.shape[-1]), device=device)], dim=1)
    theta = theta.reshape(-1, 2, 3)
    theta = F.pad(theta, (0, 0, 0, 1))
    if exp:
        theta = torch.matrix_exp(theta)
    else:
        theta[:, -1, -1] = 1
    if global_transform is not None:
        theta = theta @ global_transform
    if inverse:
        theta = torch.inverse(theta)
    grid_t = theta.matmul(grid)
    if not return_homogenous_grid:
        grid_t = grid_t[:, :2, :]
    return (theta, grid_t)

# Create minimal test data
N, C, H, W = 2, 3, 4, 5
shape = (N, C, H, W)

# Create a small image tensor
img = torch.randn(N, C, H, W)

# Create theta_dict with affine and homography transforms
theta_affine = torch.randn(N, 6)
theta_homography = torch.randn(N, 8)
theta_dict = {
    'affine': theta_affine,
    'homography': theta_homography
}

# Create optional global_transform (3x3 transformation matrix)
global_transform = torch.eye(3).unsqueeze(0).repeat(N, 1, 1)

device = 'cpu'

ARGS = {
    'img': img,
    'theta_dict': theta_dict,
    'shape': shape,
    'device': device,
    'global_transform': global_transform
}
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 Tensor, ndim 4, dtype family float, shape (2, 3, 4, 5) — shape (2, 3, 4, 5) 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
weighted_average Ran
ran; our contract draft was wrong, not the code
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("2cc6784d503beafa")
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 1, dtype family float, shape (4,) 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)

# Create tiny tensors for the weighted_average function
# mu: tensor that will be filled with weighted averages
mu = torch.zeros(4, dtype=torch.float32)

# image_acc_sum: accumulated image values
image_acc_sum = torch.tensor([1.0, 2.0, 3.0, 4.0], dtype=torch.float32)

# mask_acc_sum: accumulated mask values (for weighting)
mask_acc_sum = torch.tensor([2.0, 4.0, 0.5e-6, 8.0], dtype=torch.float32)

# zero_sensitivity: threshold for considering a value valid
zero_sensitivity = 1e-05

ARGS = {
    "mu": mu,
    "image_acc_sum": image_acc_sum,
    "mask_acc_sum": mask_acc_sum,
    "zero_sensitivity": zero_sensitivity
}
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 Tensor, ndim 1, dtype family float, shape (4,) — shape (4,) 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
BMN Not yet run
unverified (harvested, no recorded run)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("c4a327e70cd8548f")
How it ran

no invocation recorded.

trim_pixel_stack_ Not yet run
unverified (harvested, no recorded run)
BGU-CS-VIL/DeepMCBM/src/BMN.py
code served (permissive licence) · get_code("fac5bd4846e7a1b5")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 10 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.

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

In video analysis, background models have many applications such as background/foreground separation, change detection, anomaly detection, tracking, and more. However, while learning such a model in a video captured by a static camera is a fairly-solved task, in the case of a Moving-camera Background Model (MCBM), the success has been far more modest due to algorithmic and scalability challenges that arise due to the camera motion. Thus, existing MCBMs are limited in their scope and their supported camera-motion types. These hurdles also impeded the employment, in this unsupervised task, of end-to-end solutions based on deep learning (DL). Moreover, existing MCBMs usually model the background either on the domain of a typically-large panoramic image or in an online fashion. Unfortunately, the former creates several problems, including poor scalability, while the latter prevents the recognition and leveraging of cases where the camera revisits previously-seen parts of the scene. This paper proposes a new method, called DeepMCBM, that eliminates all the aforementioned issues and achieves state-of-the-art results. Concretely, first we identify the difficulties associated with joint alignment of video frames in general and in a DL setting in particular. Next, we propose a new strategy for joint alignment that lets us use a spatial transformer net with neither a regularization nor any form of specialized (and non-differentiable) initialization. Coupled with an autoencoder conditioned on unwarped robust central moments (obtained from the joint alignment), this yields an end-to-end regularization-free MCBM that supports a broad range of camera motions and scales gracefully. We demonstrate DeepMCBM's utility on a variety of videos, including ones beyond the scope of other methods. Our code is available at https://github.com/BGU-CS-VIL/DeepMCBM.

For agents

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

get_harvested_code_for_paper("2209.07923")
get_code_for_paper("2209.07923")
have("2209.07923")

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

curl https://syntology.ai/api/ran/2209.07923.json

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

[![Syntology run record](https://syntology.ai/api/ran/2209.07923.svg)](https://syntology.ai/paper/2209.07923)

Connect an agent — have() is free.