SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.00324 · CVPR · 2021

STMTrack: Template-free Visual Tracking with Space-time Memory Networks

Zehua Fu, Qingjie Liu, Yunhong Wang, Zhihong Fu

arXiv · PDF · Open in the Atlas

Code that ran

8 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 6 failures of our instrument; 3 no recorded run), measured 2026-11-22; 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/2104.00324.json · badge: /api/ran/2104.00324.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
fzh0917/STMTrack marked official in Papers with Code
Papers with Code archive
8 ran · 3 no recorded run
FunctionStatusWhere it lives
Registry Ran
ran on a synthesized input · constructed an object; no result was computed
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("3f950a5c3a805857")
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 an instance of its own class, Registry: the run constructed the object and did not call it, so no result was computed. 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


def _register_generic(module_dict, module_name, module):
    assert module_name not in module_dict, print(module_name, module_dict, 'defined in several script files')
    module_dict[module_name] = module


class Registry(dict):
    """
    A helper class for managing registering modules, it extends a dictionary
    and provides a register functions.

    usually declared in XXX_base.py, e.g. videoanalyst/model/backbone/backbone_base.py

    used as decorator when declaring the module:

    @some_registry.register
    def foo():
        ...

    Access of module is just like using a dictionary, eg:
        f = some_registry["foo_module"]
    """

    def __init__(self, *args, **kwargs):
        self.name = 'Registry'
        if len(args) > 0 and isinstance(args[0], str):
            name, *args = args
            self.name = name
        super(Registry, self).__init__(*args, **kwargs)

    def register(self, module):
        name = module.__name__
        _register_generic(self, name, module)
        return module


# Create a simple test module function to register
def test_module_1():
    pass

test_module_1.__name__ = 'test_module_1'

# Define ARGS for Registry constructor
ARGS = {
    'args': ('MyRegistry',),
    'kwargs': {}
}
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 Registry — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Registry 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
conv_bn_relu Ran
ran on a synthesized input · constructed an object; no result was computed
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("f2bebbb153632753")
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 an instance of its own class, conv_bn_relu: the run constructed the object and did not call it, so no result was computed. 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)

ARGS = {
    "in_channel": 3,
    "out_channel": 64,
    "stride": 1,
    "kszie": 3,
    "pad": 1,
    "has_bn": True,
    "has_relu": True,
    "bias": True,
    "groups": 1,
}
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 conv_bn_relu — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a conv_bn_relu 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
filter_reused_missing_keys Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("f05dd1f0fbc22a83")
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 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
from collections import defaultdict
from typing import List

torch.manual_seed(0)

# Create a model with parameters and buffers, including some reused modules
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1 = nn.Linear(4, 3)
        self.layer2 = nn.Linear(3, 2)
        self.register_buffer('buf1', torch.tensor([1.0, 2.0]))

model = SimpleModel()

# Create a list of missing keys that includes some that are actually loaded
# and some that are not, to test the filtering logic
keys = [
    'layer1.weight',
    'layer1.bias',
    'layer2.weight',
    'buf1',
    'nonexistent.param'
]

ARGS = {
    'model': model,
    'keys': keys
}
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 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
get_missing_parameters_message Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("afa2b03cb5dc9dbe")
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 str. 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
from collections import defaultdict

torch.manual_seed(0)

# Construct valid arguments for get_missing_parameters_message
# The function takes a list of strings (parameter names/keys)

ARGS = {
    "keys": [
        "layer1.weight",
        "layer1.bias",
        "layer2.weight",
        "layer2.bias",
        "encoder.weight",
        "encoder.bias",
        "decoder.output",
        "standalone_param"
    ]
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
get_unexpected_parameters_message Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("ea09e35f7d4ca931")
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 str. 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 collections import defaultdict
from typing import Any, Dict, Iterable, List, Tuple

# Construct valid arguments for get_unexpected_parameters_message
keys_example = [
    "layer1.weight",
    "layer1.bias",
    "layer2.weight",
    "layer2.bias",
    "backbone.conv.weight",
    "backbone.conv.bias",
    "unused_param"
]

ARGS = {
    "keys": keys_example
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
group_checkpoint_keys Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("289de134af2bd65d")
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 defaultdict. 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
from collections import defaultdict
from typing import Any, Dict, Iterable, List, Tuple

torch.manual_seed(0)

# The function takes a list of strings (parameter names/keys)
# It groups them by their prefix (everything before the last ".")
# No complex objects needed - just strings

keys = [
    "layer1.weight",
    "layer1.bias",
    "layer2.weight",
    "layer2.bias",
    "backbone.conv.weight",
    "backbone.conv.bias",
    "simple_param",
    "another_param"
]

ARGS = {
    "keys": keys
}
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 defaultdict — a defaultdict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a defaultdict 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
md5sum Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("df35e234671d8855")
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 str. 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 tempfile
import os

# Create a temporary file with some content
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_file.txt")
with open(temp_file, 'w') as f:
    f.write("test content for md5sum calculation")

ARGS = {
    "file_path": temp_file
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
named_modules_with_dup Ran
ran; our contract draft was wrong, not the code
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("07588af5a652bf1f")
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 generator. 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)

# Create a simple model with nested modules to test named_modules_with_dup
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.relu = nn.ReLU()
        self.linear2 = nn.Linear(3, 2)
        # Add a None module to test the None check
        self._modules['null_module'] = None
        # Create a nested submodule
        self.submodule = nn.Sequential(
            nn.Linear(2, 5),
            nn.ReLU()
        )

model = SimpleModel()

ARGS = {
    'model': model,
    'prefix': ''
}
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 generator — a generator output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a generator 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
ModuleBase Not yet run
unverified (harvested, no recorded run)
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("aafaacd38dc55bcb")
How it ran

no invocation recorded.

STMTrack Not yet run
unverified (harvested, no recorded run)
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("ccee80e86f38a6bc")
How it ran

no invocation recorded.

_register_generic Not yet run
unverified (harvested, no recorded run)
fzh0917/STMTrack/videoanalyst/model/task_model/taskmodel_impl/stmtrack_model.py
code served (permissive licence) · get_code("42a153ed65fc5b82")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 9 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 9 later) and 2 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
ModuleBase
brick of this paper's assembly: helper
1 other paper
1 later
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
Registry
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
_register_generic
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
filter_reused_missing_keys
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
get_missing_parameters_message
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
get_unexpected_parameters_message
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
group_checkpoint_keys
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
md5sum
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention
named_modules_with_dup
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2302.13840 (2023-02)
Target-Aware Tracking with Long-Term Context Attention

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

Boosting performance of the offline trained siamese trackers is getting harder nowadays since the fixed information of the template cropped from the first frame has been almost thoroughly mined, but they are poorly capable of resisting target appearance changes. Existing trackers with template updating mechanisms rely on time-consuming numerical optimization and complex hand-designed strategies to achieve competitive performance, hindering them from real-time tracking and practical applications. In this paper, we propose a novel tracking framework built on top of a space-time memory network that is competent to make full use of historical information related to the target for better adapting to appearance variations during tracking. Specifically, a novel memory mechanism is introduced, which stores the historical information of the target to guide the tracker to focus on the most informative regions in the current frame. Furthermore, the pixel-level similarity computation of the memory network enables our tracker to generate much more accurate bounding boxes of the target. Extensive experiments and comparisons with many competitive trackers on challenging large-scale benchmarks, OTB-2015, TrackingNet, GOT-10k, LaSOT, UAV123, and VOT2018, show that, without bells and whistles, our tracker outperforms all previous state-of-the-art real-time methods while running at 37 FPS. The code is available at https: //github.com/fzh0917/STMTrack.

For agents

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

get_harvested_code_for_paper("2104.00324")
get_code_for_paper("2104.00324")
have("2104.00324")

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

curl https://syntology.ai/api/ran/2104.00324.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.00324.svg)](https://syntology.ai/paper/2104.00324)

Connect an agent — have() is free.