SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1904.01906 · 2019

What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis

arXiv · PDF · Open in the Atlas

Code that ran

17 of 19 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 12 have no contract verdict; 3 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/1904.01906.json · badge: /api/ran/1904.01906.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
JoegameZhou/cnnctc community (classifier label)
classifier basis not recorded on the link
7 ran · 0 no recorded run
Eom-taeseon/CV_SceneTextRecognition community (classifier label)
classifier basis not recorded on the link
5 ran · 2 no recorded run
dmitrijsk/attentionhtr community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
roatienza/deep-text-recognition-benchmark community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
a_cosine_learning_rate Ran
ran, honoured the contract we drafted
JoegameZhou/cnnctc/src/lr_schedule.py
code served (permissive licence) · get_code("316fb42e9e6ed79e")
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 float, 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 math

def a_cosine_learning_rate(current_step, base_lr, warmup_steps, decay_steps):
    base = float(current_step - warmup_steps) / float(decay_steps)
    learning_rate = (1 + math.cos(base * math.pi)) / 2 * base_lr
    return learning_rate

ARGS = {
    "current_step": 100,
    "base_lr": 0.1,
    "warmup_steps": 10,
    "decay_steps": 1000
}
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 float, 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
calcul_acc Ran
ran on a synthesized input
JoegameZhou/cnnctc/postprocess.py
code served (permissive licence) · get_code("89c58f89ba416399")
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 float, 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)

# Analyze the function:
# - labels: iterable of values to compare
# - preds: iterable of predicted values
# - Function zips them and compares equality, then divides by len(labels)
# Both need to be iterables with comparable elements

labels = [0, 1, 0, 1, 1]
preds = [0, 1, 0, 0, 1]

ARGS = {
    "labels": labels,
    "preds": preds
}
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 float, 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
change_args Ran
ran on a synthesized input
dmitrijsk/attentionhtr/datasets/process_imgur5k.py
code served (permissive licence) · get_code("7df8b90dc252b116")
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 Args. 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 argparse

torch.manual_seed(0)

class Args:
    def __init__(self):
        self.idx_to = None

args = Args()
n_img = 42

ARGS = {
    "args": args,
    "n_img": n_img
}
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 Args — a Args output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Args 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
checkImageIsValid Ran
ran on a synthesized input
Eom-taeseon/CV_SceneTextRecognition/create_lmdb_dataset.py
code served (permissive licence) · get_code("c7b92e1633b0d2da")
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 bool, 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 cv2
import numpy as np

# Create a valid image in binary format
# Create a simple 10x8 grayscale image (10 height, 8 width)
img = np.random.randint(0, 256, size=(10, 8), dtype=np.uint8)
# Encode to PNG format (or any format that cv2.imdecode can read)
success, imageBin = cv2.imencode('.png', img)
imageBin = imageBin.tobytes()

ARGS = {
    "imageBin": imageBin
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, 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
dynamic_lr Ran
ran on a synthesized input
JoegameZhou/cnnctc/src/lr_schedule.py
code served (permissive licence) · get_code("fe2c44cd738c9d3e")
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 math

class Config:
    def __init__(self):
        self.base_lr = 0.001
        self.TRAIN_EPOCHS = 5
        self.warmup_step = 10
        self.warmup_ratio = 0.1

config = Config()
steps_per_epoch = 20

ARGS = {
    "config": config,
    "steps_per_epoch": steps_per_epoch,
}
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 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_args Ran
ran on a synthesized input
Eom-taeseon/CV_SceneTextRecognition/utils.py
code served (permissive licence) · get_code("ce83d7152f32b55d")
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 Namespace. 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 argparse
import sys

ARGS = {
    'is_train': True,
}

sys.argv = [
    'script.py',
    '--train_data', '/tmp/train_data',
    '--valid_data', '/tmp/valid_data',
    '--Transformation', 'TPS',
    '--FeatureExtraction', 'ResNet',
    '--SequenceModeling', 'BiLSTM',
    '--Prediction', 'Attn',
    '--exp_name', 'test_exp',
    '--batch_size', '32',
    '--num_iter', '1000',
    '--lr', '0.1',
    '--manualSeed', '42',
    '--workers', '2',
    '--valInterval', '500',
    '--imgH', '32',
    '--imgW', '100',
    '--batch_max_length', '25',
    '--hidden_size', '256',
    '--output_channel', '512',
    '--input_channel', '1',
    '--num_fiducial', '20',
    '--character', '0123456789abcdefghijklmnopqrstuvwxyz',
    '--select_data', 'MJ-ST',
    '--batch_ratio', '0.5-0.5',
    '--total_data_usage_ratio', '1.0',
    '--TransformerModel', 'vitstr_tiny_patch16_224',
]
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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_args Ran
ran on a synthesized input
roatienza/deep-text-recognition-benchmark/utils.py
code served (permissive licence) · get_code("98206309c9298631")
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 Namespace. 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 argparse
import sys

# Since get_args uses sys.argv, we need to mock the command line arguments
sys.argv = ['program']

# Define ARGS dictionary with all required parameters for get_args
ARGS = {
    'is_train': True
}

# The function get_args() will be called with is_train=True
# This means it will require --train_data and --valid_data but not --eval_data
# We need to set up sys.argv with appropriate command line arguments

sys.argv.extend([
    '--train_data', '/path/to/train_data',
    '--valid_data', '/path/to/valid_data',
    '--Transformation', 'TPS',
    '--FeatureExtraction', 'ResNet',
    '--SequenceModeling', 'BiLSTM',
    '--Prediction', 'Attn',
    '--exp_name', 'test_exp',
    '--manualSeed', '42',
    '--workers', '4',
    '--batch_size', '32',
    '--num_iter', '1000',
    '--valInterval', '100',
    '--lr', '0.1',
    '--beta1', '0.9',
    '--rho', '0.95',
    '--eps', '1e-8',
    '--grad_clip', '5.0',
    '--select_data', 'MJ-ST',
    '--batch_ratio', '0.5-0.5',
    '--total_data_usage_ratio', '1.0',
    '--batch_max_length', '25',
    '--imgH', '32',
    '--imgW', '100',
    '--character', '0123456789abcdefghijklmnopqrstuvwxyz',
    '--TransformerModel', 'vitstr_tiny_patch16_224',
    '--num_fiducial', '20',
    '--input_channel', '1',
    '--output_channel', '512',
    '--hidden_size', '256',
])
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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_device Ran
ran on a synthesized input
Eom-taeseon/CV_SceneTextRecognition/utils.py
code served (permissive licence) · get_code("291192e39061a450")
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 device. 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)

ARGS = {
    'verbose': True
}
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 device — a device output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a device 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_image_split Ran
ran on a synthesized input
dmitrijsk/attentionhtr/datasets/process_imgur5k.py
code served (permissive licence) · get_code("55becdbd1817adde")
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 numpy as np

torch.manual_seed(0)

SPLITS = ['train', 'val', 'test']

# Construct split_idx dictionary
# This maps each split name to a set/list of image names
split_idx = {
    'train': ['image_001.jpg', 'image_002.jpg', 'image_003.jpg'],
    'val': ['image_004.jpg', 'image_005.jpg'],
    'test': ['image_006.jpg', 'image_007.jpg']
}

# Test image name that exists in one of the splits
img_name = 'image_002.jpg'

ARGS = {
    'img_name': img_name,
    'split_idx': split_idx
}
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 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
img2text Ran
ran on a synthesized input
roatienza/deep-text-recognition-benchmark/infer.py
code served (permissive licence) · get_code("d8e30cd1ae314959")
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

torch.manual_seed(0)

# Build a minimal model that accepts img and seqlen parameter
class TextRecognitionModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(4, 50)  # input dim 4, output 50 (vocab size)
    
    def forward(self, img, seqlen):
        # img shape: (1, 4) -> output (1, seqlen, 50)
        batch_size = img.shape[0]
        x = self.fc(img)  # (1, 50)
        # Expand to (1, seqlen, 50)
        output = x.unsqueeze(1).expand(batch_size, seqlen, -1)
        return output

# Build a minimal converter with required attributes and methods
class Converter:
    def __init__(self):
        self.batch_max_length = 5
    
    def decode(self, pred_index, length_for_pred):
        # pred_index shape: (1, 4) after slicing [:, 1:]
        # Return list with one decoded string
        return ["hello[s]world"]

# Create instances
model = TextRecognitionModel()
converter = Converter()

# Create list of small image tensors (batch size 1, 4 features each)
images = [
    torch.randn(1, 4),
    torch.randn(1, 4)
]

ARGS = {
    "model": model,
    "images": images,
    "converter": converter
}
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
isless Ran
ran on a synthesized input
Eom-taeseon/CV_SceneTextRecognition/dataset.py
code served (permissive licence) · get_code("8f99ed40edd07e1e")
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 bool, 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 numpy as np

ARGS = {
    "prob": 0.5
}
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 bool, 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
linear_warmup_learning_rate Ran
ran, honoured the contract we drafted
JoegameZhou/cnnctc/src/lr_schedule.py
code served (permissive licence) · get_code("d6a99ec714147ac1")
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 float, 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)

# Analyzing the function:
# - current_step: used in arithmetic (int or float)
# - warmup_steps: used in division (int or float)
# - base_lr: used in float() conversion and subtraction (int or float)
# - init_lr: used in float() conversion and addition (int or float)

ARGS = {
    "current_step": 500,
    "warmup_steps": 1000,
    "base_lr": 0.1,
    "init_lr": 0.001,
}
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 float, 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
merge Ran
ran; our contract draft was wrong, not the code
JoegameZhou/cnnctc/src/model_utils/config.py
code served (permissive licence) · get_code("b2108b927c53418b")
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 torch
import argparse

torch.manual_seed(0)

class SimpleArgs:
    def __init__(self):
        self.learning_rate = 0.001
        self.batch_size = 32
        self.epochs = 10
        self.model_name = "resnet50"
        self.seed = 42
        self.debug = True

args = SimpleArgs()

cfg = {
    "task": "classification",
    "dataset": "cifar10",
    "num_classes": 10,
    "hidden_dim": 128
}

ARGS = {
    "args": args,
    "cfg": cfg
}
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 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
parse_cli_to_yaml Ran
ran; our contract draft was wrong, not the code
JoegameZhou/cnnctc/src/model_utils/config.py
code served (permissive licence) · get_code("3498f707d22460cd")
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 Namespace. 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 argparse
import ast

# Create minimal test arguments for parse_cli_to_yaml

# Base configuration dict with various types
cfg = {
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 100,
    'use_cuda': False,
    'model_name': 'resnet50',
    'dropout': 0.5,
    'nested_list': [1, 2, 3],  # Should be skipped (list type)
    'nested_dict': {'key': 'value'},  # Should be skipped (dict type)
}

# Helper descriptions for some parameters
helper = {
    'learning_rate': 'Learning rate for optimizer',
    'batch_size': 'Batch size for training',
    'use_cuda': 'Whether to use CUDA',
    'model_name': 'Name of the model to use',
}

# Choices for some parameters
choices = {
    'model_name': ['resnet50', 'resnet101', 'vgg16'],
}

# Parent parser (empty in this case)
parser = argparse.ArgumentParser(add_help=False)

# Path to config file
cfg_path = 'default_config.yaml'

ARGS = {
    'parser': parser,
    'cfg': cfg,
    'helper': helper,
    'choices': choices,
    'cfg_path': cfg_path,
}
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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
parse_yaml Ran
ran; our contract draft was wrong, not the code
JoegameZhou/cnnctc/src/model_utils/config.py
code served (permissive licence) · get_code("c8d98ae32946f6f0")
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 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 tempfile
import os
import yaml

# Create a temporary YAML file with valid config structure
temp_dir = tempfile.mkdtemp()
yaml_file = os.path.join(temp_dir, "test_config.yaml")

# Write a valid YAML file with 3 documents (config, helper, choices)
yaml_content = """
model: resnet50
learning_rate: 0.001
batch_size: 32
---
model: "The model architecture to use"
learning_rate: "Learning rate for training"
batch_size: "Batch size for training"
---
model: [resnet50, vgg16, mobilenet]
learning_rate: [0.0001, 0.001, 0.01]
batch_size: [16, 32, 64]
"""

with open(yaml_file, 'w') as f:
    f.write(yaml_content)

ARGS = {
    "yaml_path": yaml_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — 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 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
resize Ran
ran on a synthesized input
dmitrijsk/attentionhtr/datasets/process_imgur5k.py
code served (permissive licence) · get_code("e7cc5ea8a6820dc3")
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 int, shape (32, 24, 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 cv2
import numpy as np

# Create a small test image with shape (height, width, channels)
# Using a simple numpy array with distinct dimensions for clarity
img = np.random.randint(0, 256, size=(64, 48, 3), dtype=np.uint8)

ARGS = {
    "img": img,
    "desired_height": 32
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 int, shape (32, 24, 3) — shape (32, 24, 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
tensor2im Ran
ran on a synthesized input
Eom-taeseon/CV_SceneTextRecognition/dataset.py
code served (permissive licence) · get_code("b78e81e855133e45")
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 int, 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 torch
import numpy as np

torch.manual_seed(0)

# Create a small image tensor with shape (C, H, W)
# Using shape (1, 4, 3) - 1 channel, 4 height, 3 width
# Values in range [-1, 1] as expected by the function
image_tensor = torch.randn(1, 4, 3)

ARGS = {
    "image_tensor": image_tensor,
    "imtype": np.uint8
}
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 ndarray, ndim 3, dtype family int, 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
create_vitstr Not yet run
unverified (harvested, no recorded run)
Eom-taeseon/CV_SceneTextRecognition/modules/vitstr.py
code served (permissive licence) · get_code("8a4fd52ddbb2af6b")
How it ran

no invocation recorded.

get_flops Not yet run
unverified (harvested, no recorded run)
Eom-taeseon/CV_SceneTextRecognition/predict.py
code served (permissive licence) · get_code("edfd158c1fa7f00e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 distinct functions with an identity, 9 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 0 the same month, 9 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.

FunctionAlso present inEarliest-datedOthers named
merge 28 other papers
12 dated earlier, 16 later
1406.1078 (2014-06)
Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
1409.3215 (2014-09)
Sequence to Sequence Learning with Neural Networks
1512.00567 (2015-12)
Rethinking the Inception Architecture for Computer Vision
and 25 more not named here
parse_cli_to_yaml 20 other papers
10 dated earlier, 10 later
1406.1078 (2014-06)
Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
1409.3215 (2014-09)
Sequence to Sequence Learning with Neural Networks
1512.00567 (2015-12)
Rethinking the Inception Architecture for Computer Vision
and 17 more not named here
linear_warmup_learning_rate 7 other papers
3 dated earlier, 4 later
1506.01497 (2015-06)
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
1606.04797 (2016-06)
V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
1606.06650 (2016-06)
3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation
and 4 more not named here
a_cosine_learning_rate 4 other papers
2 dated earlier, 2 later
1506.01497 (2015-06)
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
1606.06650 (2016-06)
3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation
2103.13413 (2021-03)
Vision Transformers for Dense Prediction
and 1 more not named here
parse_yaml 4 other papers
1 dated earlier, 3 later
1804.06300 (2018-04)
PredRNN++: Towards A Resolution of the Deep-in-Time Dilemma in Spatiotemporal Predictive Learning
2006.04388 (2020-06)
Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection
2103.13413 (2021-03)
Vision Transformers for Dense Prediction
and 1 more not named here
tensor2im 4 other papers
1 dated earlier, 3 later
1811.00751 (2018-11)
Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition
1906.01502 (2019-06)
How multilingual is Multilingual BERT?
2207.04675 (2022-07)
COO: Comic Onomatopoeia Dataset for Recognizing Arbitrary or Truncated Texts
and 1 more not named here
checkImageIsValid 3 other papers
1 dated earlier, 2 later
1811.00751 (2018-11)
Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition
2207.04675 (2022-07)
COO: Comic Onomatopoeia Dataset for Recognizing Arbitrary or Truncated Texts
2305.05140 (2023-05)
Linguistic More: Taking a Further Step toward Efficient and Accurate Scene Text Recognition
get_device 1 other paper
1 later
2305.05140 (2023-05)
Linguistic More: Taking a Further Step toward Efficient and Accurate Scene Text Recognition
isless 1 other paper
1 later
2305.05140 (2023-05)
Linguistic More: Taking a Further Step toward Efficient and Accurate Scene Text Recognition

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

Many new proposals for scene text recognition (STR) models have been introduced in recent years. While each claim to have pushed the boundary of the technology, a holistic and fair comparison has been largely missing in the field due to the inconsistent choices of training and evaluation datasets. This paper addresses this difficulty with three major contributions. First, we examine the inconsistencies of training and evaluation datasets, and the performance gap results from inconsistencies. Second, we introduce a unified four-stage STR framework that most existing STR models fit into. Using this framework allows for the extensive evaluation of previously proposed STR modules and the discovery of previously unexplored module combinations. Third, we analyze the module-wise contributions to performance in terms of accuracy, speed, and memory demand, under one consistent set of training and evaluation datasets. Such analyses clean up the hindrance on the current comparisons to understand the performance gain of the existing modules.

For agents

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

get_harvested_code_for_paper("1904.01906")
get_code_for_paper("1904.01906")
have("1904.01906")

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

curl https://syntology.ai/api/ran/1904.01906.json

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

[![Syntology run record](https://syntology.ai/api/ran/1904.01906.svg)](https://syntology.ai/paper/1904.01906)

Connect an agent — have() is free.