SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.03673 · ICCV · 2021

RECALL: Replay-based Continual Learning in Semantic Segmentation

Pietro Zanuttigh, Umberto Michieli, Andrea Maracani, Marco Toldo

arXiv · PDF · Open in the Atlas

Code that ran

13 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 2 failures of our instrument; 5 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/2108.03673.json · badge: /api/ran/2108.03673.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
lttm/recall canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
LTTM/RECALL community
Papers with Code archive
5 ran · 4 no recorded run
FunctionStatusWhere it lives
Model Ran
ran on a synthesized input · constructed an object; no result was computed
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("a3f7331bbdb98e59")
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, Model: 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 tensorflow as tf
import numpy as np

# Enable eager execution for easier testing
tf.compat.v1.enable_eager_execution()

ARGS = {
    'parameters': {
        'helper_size': 128,
        'n_classes': 10,
        'dataset_mean': [0.485, 0.456, 0.406],
        'helper_decoder_one_class': True,
        'momentum': 0.9,
        'start_decay_step': 1000,
        'decay_steps_per_class': 5000,
        'decay_steps_per_class_incremental': 3000,
        'decay_steps_per_class_helper': 2000,
        'decay_power': 0.9,
        'l2_decay': 1e-4,
        'lr_initial_value': 0.01,
        'lr_final_value': 0.001,
        'lr_initial_value_incremental': 0.005,
        'lr_final_value_incremental': 0.0005,
        'lr_initial_value_helper': 0.01,
        'lr_final_value_helper': 0.001,
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, six, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Model — 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 Model 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
add Ran
ran, honoured the contract we drafted
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("a96ff10f26036567")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Create a list of TensorFlow tensors to add
tensor_1 = tf.constant([1.0, 2.0, 3.0, 4.0], dtype=tf.float32)
tensor_2 = tf.constant([5.0, 6.0, 7.0, 8.0], dtype=tf.float32)
tensor_3 = tf.constant([9.0, 10.0, 11.0, 12.0], dtype=tf.float32)

x_l = [tensor_1, tensor_2, tensor_3]
name = "test_add"

ARGS = {
    "x_l": x_l,
    "name": name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
get_classes_accuracy Ran
ran on a synthesized input
lttm/recall/metrics.py
code served (permissive licence) · get_code("92fa3b85dd6d9930")
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 numpy as np

torch.manual_seed(0)

# Create a small confusion matrix
# Shape: (n_classes, n_classes) = (4, 4)
# For a confusion matrix, rows are true labels, columns are predicted labels
# The diagonal elements are correct predictions
confusion_matrix = np.array([
    [10, 2, 1, 0],
    [1, 8, 1, 0],
    [0, 1, 6, 2],
    [0, 0, 1, 5]
], dtype=np.float32)

# Classes to evaluate (subset of available classes)
classes = [0, 1, 2]

ARGS = {
    "confusion_matrix": confusion_matrix,
    "classes": classes
}
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 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_classes_iou Ran
ran on a synthesized input
lttm/recall/metrics.py
code served (permissive licence) · get_code("8e11a0aa76c799b6")
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 numpy as np

torch.manual_seed(0)

# Create a small confusion matrix
# Shape: (4, 4) - 4 classes
# Values represent pixel counts in a confusion matrix
confusion_matrix = np.array([
    [50, 5, 2, 1],
    [3, 60, 4, 2],
    [1, 2, 55, 3],
    [2, 1, 3, 48]
], dtype=np.float32)

# Define classes to compute IoU for
classes = [0, 1, 2, 3]

ARGS = {
    "confusion_matrix": confusion_matrix,
    "classes": classes
}
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 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_mean_pixel_accuracy Ran
ran on a synthesized input
lttm/recall/metrics.py
code served (permissive licence) · get_code("48bc91b33344e330")
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)

# Build a small confusion matrix
# Shape: (n_classes, n_classes) where rows are true labels, columns are predicted labels
# Using 4x4 for a small example with 4 classes
confusion_matrix = torch.tensor([
    [10, 2, 1, 0],
    [1, 15, 2, 1],
    [0, 1, 12, 2],
    [1, 0, 1, 14]
], dtype=torch.float32)

# Classes to evaluate - subset of available classes
classes = [0, 2, 3]

ARGS = {
    "confusion_matrix": confusion_matrix,
    "classes": classes
}
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
im2int Ran
ran on a synthesized input
lttm/recall/gan/generate_gan.py
code served (permissive licence) · get_code("7045941346cccb6f")
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 EagerTensor, ndim 3, dtype family int, shape (2, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# The function im2int expects an image tensor that will be:
# 1. Added with 1
# 2. Divided by 2
# 3. Converted to uint8 dtype
# 
# A typical image input would be a tensor with values in range [-1, 1]
# (normalized image). After (image + 1) / 2, values will be in [0, 1]
# range, which is appropriate for uint8 conversion.

# Create a small test image tensor
# Using shape (2, 4, 3) for batch_size=2, height=4, width=3, channels=3
image_data = np.random.uniform(-1, 1, size=(2, 4, 3)).astype(np.float32)
image_tensor = tf.constant(image_data)

ARGS = {
    "image": image_tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 3, dtype family int, shape (2, 4, 3) — shape (2, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
interleave_datasets Ran
ran on a synthesized input
lttm/recall/data_loader.py
code served (permissive licence) · get_code("e74234eaa6937655")
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 _BatchDataset. 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 tensorflow as tf

# Create simple tf.data.Dataset objects with dictionaries containing 'input' and 'label' keys
def create_sample_dataset(num_samples=4, input_dim=4, label_dim=1):
    inputs = tf.constant([[i, i+1, i+2, i+3] for i in range(num_samples)], dtype=tf.float32)
    labels = tf.constant([[i] for i in range(num_samples)], dtype=tf.float32)
    
    # Create dataset of dictionaries
    dataset = tf.data.Dataset.from_tensor_slices({
        'input': inputs,
        'label': labels
    }).batch(2)
    
    return dataset

dataset1 = create_sample_dataset(num_samples=4, input_dim=4, label_dim=1)
dataset2 = create_sample_dataset(num_samples=4, input_dim=4, label_dim=1)

ARGS = {
    'dataset1': dataset1,
    'dataset2': dataset2,
    'r1': 2,
    'r2': 2,
    'batch_size': 2,
    'buffer_size': 100,
    'repeat': True,
    'shuffle': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type _BatchDataset — a _BatchDataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a _BatchDataset 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
max_pool2d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("ded213268c033a96")
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 EagerTensor, ndim 4, dtype family float, shape (2, 4, 3, 5) 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 tensorflow as tf
import numpy as np

# Create a small input tensor for max_pool2d
# The function expects x to be a 4D tensor in NHWC format (batch, height, width, channels)
# Using batch_size=2, height=4, width=3, channels=5 for distinct dimensions
x = tf.constant(np.random.randn(2, 4, 3, 5).astype(np.float32))

# kernel_size: integer specifying the pooling window size
kernel_size = 2

# stride: integer specifying the stride of the pooling operation
stride = 1

# name: string identifier for the operation
name = "test_max_pool"

ARGS = {
    "x": x,
    "kernel_size": kernel_size,
    "stride": stride,
    "name": name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 4, dtype family float, shape (2, 4, 3, 5) — shape (2, 4, 3, 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
preprocess_validation_op Ran
ran on a synthesized input
lttm/recall/preprocessing.py
code served (permissive licence) · get_code("94725ecbda96c654")
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. 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 tensorflow as tf
import numpy as np

# Create tiny test tensors
image = tf.constant(np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8), dtype=tf.uint8)
label = tf.constant(np.random.randint(0, 10, size=(4, 3, 1), dtype=np.int32), dtype=tf.int32)

# Create a mean tensor with same spatial dimensions as image
mean = tf.constant(np.ones((4, 3, 3), dtype=np.float32) * 127.5, dtype=tf.float32)

ARGS = {
    "image": image,
    "label": label,
    "subtract_mean": True,
    "mean": mean
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — 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
relu Ran
ran, honoured the contract we drafted
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("37e49142da9ee0cd")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Create a small TensorFlow tensor for input
x = tf.constant([[1.0, -2.0, 3.0, -4.0],
                 [5.0, -6.0, 7.0, -8.0]], dtype=tf.float32)

# Define arguments for the relu function
ARGS = {
    "x": x,
    "name": "test_relu"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
resize_image_op Ran
ran on a synthesized input
lttm/recall/preprocessing.py
code served (permissive licence) · get_code("423a14145bc09771")
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 EagerTensor, ndim 3, dtype family float, shape (8, 8, 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 tensorflow as tf
import numpy as np

# Create a small test image: 3D tensor [width, height, channels]
# Using shape [4, 5, 3] for a 4x5 RGB image
image = tf.constant(np.random.randint(0, 256, (4, 5, 3), dtype=np.uint8), dtype=tf.uint8)

# Size parameter - output image will be resized to [size, size]
size = 8

# subtract_mean parameter
subtract_mean = True

# mean parameter - a tensor to subtract from the image
# Should have shape compatible with the image for broadcasting
mean = tf.constant(np.random.rand(1, 1, 3).astype(np.float32) * 50, dtype=tf.float32)

ARGS = {
    "image": image,
    "size": size,
    "subtract_mean": subtract_mean,
    "mean": mean
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 3, dtype family float, shape (8, 8, 3) — shape (8, 8, 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
serialize_sample Ran
ran on a synthesized input
lttm/recall/data_loader.py
code served (permissive licence) · get_code("2e9c72ac0591430c")
How it ran

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

import tensorflow as tf
import numpy as np

# Construct arguments for serialize_sample function
# The function expects:
# - image: something that can be converted to bytes via _bytes_feature
# - label: something that can be converted to bytes via _bytes_feature

# Create simple image data as bytes
image_data = b"fake_image_data_123"

# Create simple label data as bytes
label_data = b"class_0"

ARGS = {
    "image": image_data,
    "label": label_data
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bytes — a bytes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a bytes output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
upsampling Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("5d9c5fe9e4c597a9")
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 EagerTensor, ndim 4, dtype family float, shape (2, 8, 10, 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 tensorflow as tf
import numpy as np

# Create a small input tensor for upsampling
# Shape: (batch_size, height, width, channels) = (2, 4, 5, 3)
inputs = tf.constant(np.random.randn(2, 4, 5, 3).astype(np.float32))

# Define output_shape as a tuple/list with format (batch, height, width, channels)
# The function uses output_shape[1] and output_shape[2] for height and width
output_shape = (2, 8, 10, 3)

ARGS = {
    "inputs": inputs,
    "output_shape": output_shape,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 4, dtype family float, shape (2, 8, 10, 3) — shape (2, 8, 10, 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
ASPP Not yet run
unverified (harvested, no recorded run)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("7130868dd7f4080f")
How it ran

no invocation recorded.

batch_norm Not yet run
unverified (harvested, no recorded run)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("54d981e50a6f3e0f")
How it ran

no invocation recorded.

conv2d Not yet run
unverified (harvested, no recorded run)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("22aa81d960081ea7")
How it ran

no invocation recorded.

dilated_conv2d Not yet run
unverified (harvested, no recorded run)
LTTM/RECALL/model_resnet.py
code served (permissive licence) · get_code("391774518dbd2cc0")
How it ran

no invocation recorded.

preprocess_training_op Not yet run
unverified (harvested, no recorded run)
lttm/recall/preprocessing.py
code served (permissive licence) · get_code("3cdee8d3ea4218a4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 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 18 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

Deep networks allow to obtain outstanding results in semantic segmentation, however they need to be trained in a single shot with a large amount of data. Continual learning settings where new classes are learned in incremental steps and previous training data is no longer available are challenging due to the catastrophic forgetting phenomenon. Existing approaches typically fail when several incremental steps are performed or in presence of a distribution shift of the background class. We tackle these issues by recreating no longer available data for the old classes and outlining a content inpainting scheme on the background class. We propose two sources for replay data. The first resorts to a generative adversarial network to sample from the class space of past learning steps. The second relies on web-crawled data to retrieve images containing examples of old classes from online databases. In both scenarios no samples of past steps are stored, thus avoiding privacy concerns. Replay data are then blended with new samples during the incremental steps. Our approach, RECALL, outperforms state-of-the-art methods.

For agents

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

get_harvested_code_for_paper("2108.03673")
get_code_for_paper("2108.03673")
have("2108.03673")

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

curl https://syntology.ai/api/ran/2108.03673.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.03673.svg)](https://syntology.ai/paper/2108.03673)

Connect an agent — have() is free.