SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2008.11516 · 2020

Making a Case for 3D Convolutions for Object Segmentation in Videos

arXiv · PDF · Open in the Atlas

Code that ran

9 of 9 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 6 have no contract verdict; 3 failures of our instrument; 0 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/2008.11516.json · badge: /api/ran/2008.11516.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
sabarim/3DC-Seg canonical (classifier label)
classifier basis: census
9 ran · 0 no recorded run
FunctionStatusWhere it lives
blobs_from_pkl Ran
ran on a synthesized input
sabarim/3DC-Seg/network/convert_csn_weights.py
code served (permissive licence) · get_code("3b84b83a1ad1cb0f")
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 dict. 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 pickle
import tempfile
from pathlib import Path

# Create a temporary pickle file with the expected structure
temp_dir = tempfile.TemporaryDirectory()
temp_path = Path(temp_dir.name) / "test.pkl"

# Create blobs dict with the required keys for num_classes=400
blobs = {
    'last_out_L400_w': [0.1, 0.2, 0.3],
    'last_out_L400_b': [0.05, 0.15],
    'other_data': [1, 2, 3, 4, 5]
}

pkl_data = {'blobs': blobs}

with temp_path.open(mode='wb') as f:
    pickle.dump(pkl_data, f)

ARGS = {
    'path': temp_path,
    'num_classes': 400
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
bootstrapped_ce_loss Ran
ran on a synthesized input
sabarim/3DC-Seg/loss/loss_utils.py
code served (permissive licence) · get_code("9e3beec0999a954f")
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 0, dtype family float, shape () 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)

# Create a raw_ce tensor with shape (batch, channels, height, width)
# Using distinct dimensions: batch=2, channels=3, height=4, width=5
raw_ce = torch.randn(2, 3, 4, 5)

# n_valid_pixels_per_im: number of valid pixels per image
# Can be an integer or None (will default to height * width)
n_valid_pixels_per_im = 20

# fraction: fraction of pixels to select via topk
fraction = 0.25

ARGS = {
    "raw_ce": raw_ce,
    "n_valid_pixels_per_im": n_valid_pixels_per_im,
    "fraction": fraction,
}
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 0, dtype family float, 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
calc_iou Ran
ran on a synthesized input
sabarim/3DC-Seg/loss/loss_utils.py
code served (permissive licence) · get_code("d1494618265be1d5")
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 0, dtype family float, 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 numpy as np

torch.manual_seed(0)

# The function expects:
# - pred: a tensor with shape (num_frames, num_classes, height, width)
#   It calls torch.argmax(pred, dim=1) so it needs at least 2 dims
#   and a class dimension at dim=1
# - gt: a tensor with shape (num_frames, height, width)
#   Ground truth labels
# - info: optional parameter, can be None

num_frames = 2
num_classes = 4
height = 3
width = 5

# Create pred tensor: (num_frames, num_classes, height, width)
# Contains logits/scores for each class
pred = torch.randn(num_frames, num_classes, height, width)

# Create gt tensor: (num_frames, height, width)
# Contains class labels (0 for background, 1+ for objects)
gt = torch.randint(0, num_classes, (num_frames, height, width))

ARGS = {
    "pred": pred,
    "gt": gt,
    "info": None
}
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 Tensor, ndim 0, dtype family float, 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
conv3x3x3 Ran
ran; our contract draft was wrong, not the code
sabarim/3DC-Seg/network/Resnet3d.py
code served (permissive licence) · get_code("c3b215555357b9d1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Conv3d. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# The function conv3x3x3 creates a 3D convolution layer
# Parameters:
# - in_planes: number of input channels
# - out_planes: number of output channels  
# - stride: stride for the convolution (default=1)

# We need to provide distinct values for in_planes and out_planes
in_planes = 3
out_planes = 5
stride = 2

ARGS = {
    'in_planes': in_planes,
    'out_planes': out_planes,
    'stride': stride
}
Setup block drafted by
a language model, Claude Sonnet 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 Conv3d — a Conv3d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv3d 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
downsample_basic_block Ran
ran; our contract draft was wrong, not the code
sabarim/3DC-Seg/network/Resnet3d.py
code served (permissive licence) · get_code("43c0211ff8dbca5f")
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 Tensor, ndim 5, dtype family float, shape (2, 8, 2, 2, 2) 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
from torch.autograd import Variable

torch.manual_seed(0)

# Create a small 5D tensor (batch_size=2, channels=4, depth=3, height=3, width=3)
# This matches the expected input for a 3D convolution operation
x = torch.randn(2, 4, 3, 3, 3)

# planes should be greater than the number of input channels (4)
# so that zero padding can be added
planes = 8

# stride for the pooling operation
stride = 2

ARGS = {
    "x": x,
    "planes": planes,
    "stride": stride
}
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 Tensor, ndim 5, dtype family float, shape (2, 8, 2, 2, 2) — shape (2, 8, 2, 2, 2) 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_fine_tuning_parameters Ran
ran; our contract draft was wrong, not the code
sabarim/3DC-Seg/network/Resnet3d.py
code served (permissive licence) · get_code("8d57521c7b52b777")
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 list, ndim 1, dtype family float; 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)

# Build a minimal model that has named_parameters with layer0-4 and fc
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer0 = nn.Linear(4, 3)
        self.layer1 = nn.Linear(3, 3)
        self.layer2 = nn.Linear(3, 3)
        self.layer3 = nn.Linear(3, 3)
        self.layer4 = nn.Linear(3, 2)
        self.fc = nn.Linear(2, 1)
    
    def forward(self, x):
        x = self.layer0(x)
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)
        x = self.fc(x)
        return x

model = SimpleModel()

ARGS = {
    'model': model,
    'ft_begin_index': 2
}
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 list, ndim 1, dtype family float — a list 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
propagate3d Ran
ran on a synthesized input
sabarim/3DC-Seg/network/NetworkUtil.py
code served (permissive licence) · get_code("9cb93dd46a39dbcd")
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 2, dtype family float, shape (2, 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)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, inputs, ref_mask):
        # inputs shape: (batch, channels, time, ...)
        # ref_mask shape: (batch, ...)
        # return something with same batch/channel structure
        batch_size = inputs.shape[0]
        return torch.randn(batch_size, 4)

model = SimpleModel()

# inputs: (batch=2, channels=4, time=3, spatial=5)
# Must have time dimension >= 2 (checked in assert inputs.shape[2] >= 2)
inputs = torch.randn(2, 4, 3, 5)

# ref_mask: (batch=2, spatial=5)
ref_mask = torch.randn(2, 5)

# proposals: not used in function body, can be any value
proposals = None

ARGS = {
    "model": model,
    "inputs": inputs,
    "ref_mask": ref_mask,
    "proposals": proposals
}
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 2, dtype family float, shape (2, 4) — shape (2, 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
r2plus1d_34 Ran
ran on a synthesized input
sabarim/3DC-Seg/network/R2plus1d.py
code served (permissive licence) · get_code("e227a1a6652f7f3d")
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 VideoResNet. 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 torch
from torch import nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10,
    'pretrained': False,
    'progress': False,
    'arch': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type VideoResNet — a VideoResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a VideoResNet 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
run_forward Ran
ran on a synthesized input
sabarim/3DC-Seg/network/NetworkUtil.py
code served (permissive licence) · get_code("58780db97ad295fe")
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 2, dtype family float, shape (2, 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)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, inputs, ref_mask):
        # inputs: [batch, channels, depth, ...]
        # ref_mask: [batch, ...]
        # Return a tensor with same batch size as inputs
        batch_size = inputs.shape[0]
        return torch.randn(batch_size, 4)

model = SimpleModel()

# inputs must have shape[2] >= 2 (depth dimension >= 2)
inputs = torch.randn(2, 4, 3, 5)  # [batch=2, channels=4, depth=3, spatial=5]

# ref_masks should match inputs batch dimension
ref_masks = torch.randint(0, 2, (2, 3, 5))  # [batch=2, spatial dims]

# proposals - not used in the function body, can be any value
proposals = None

ARGS = {
    "model": model,
    "inputs": inputs,
    "ref_masks": ref_masks,
    "proposals": proposals
}
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 2, dtype family float, shape (2, 4) — shape (2, 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

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 1 the same month, 3 later) and 6 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
conv3x3x3 19 other papers
10 dated earlier, 9 later
1711.08580 (2017-11)
3D Anisotropic Hybrid Network: Transferring Convolutional Features from 2D Images to 3D Anisotropic Volumes
1711.09577 (2017-11)
Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?
1908.01373 (2019-08)
Unsupervised Microvascular Image Segmentation Using an Active Contours Mimicking Neural Network
and 16 more not named here
downsample_basic_block 15 other papers
9 dated earlier, 1 the same month, 5 later
1705.07750 (2017-05)
Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset
1711.09577 (2017-11)
Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?
1908.01373 (2019-08)
Unsupervised Microvascular Image Segmentation Using an Active Contours Mimicking Neural Network
and 12 more not named here
get_fine_tuning_parameters 3 other papers
1 dated earlier, 2 later
1711.09577 (2017-11)
Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?
2307.16184 (2023-07)
UnIVAL: Unified Model for Image, Video, Audio and Language Tasks
2311.12052 (2023-11)
MagicPose: Realistic Human Poses and Facial Expressions Retargeting with Identity-aware Diffusion

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

The task of object segmentation in videos is usually accomplished by processing appearance and motion information separately using standard 2D convolutional networks, followed by a learned fusion of the two sources of information. On the other hand, 3D convolutional networks have been successfully applied for video classification tasks, but have not been leveraged as effectively to problems involving dense per-pixel interpretation of videos compared to their 2D convolutional counterparts and lag behind the aforementioned networks in terms of performance. In this work, we show that 3D CNNs can be effectively applied to dense video prediction tasks such as salient object segmentation. We propose a simple yet effective encoder-decoder network architecture consisting entirely of 3D convolutions that can be trained end-to-end using a standard cross-entropy loss. To this end, we leverage an efficient 3D encoder, and propose a 3D decoder architecture, that comprises novel 3D Global Convolution layers and 3D Refinement modules. Our approach outperforms existing state-of-the-arts by a large margin on the DAVIS'16 Unsupervised, FBMS and ViSal dataset benchmarks in addition to being faster, thus showing that our architecture can efficiently learn expressive spatio-temporal features and produce high quality video segmentation masks. We have made our code and trained models publicly available at https://github.com/sabarim/3DC-Seg.

For agents

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

get_harvested_code_for_paper("2008.11516")
get_code_for_paper("2008.11516")
have("2008.11516")

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

curl https://syntology.ai/api/ran/2008.11516.json

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

[![Syntology run record](https://syntology.ai/api/ran/2008.11516.svg)](https://syntology.ai/paper/2008.11516)

Connect an agent — have() is free.