SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1902.04502 · 2019

Fast-SCNN: Fast Semantic Segmentation Network

arXiv · PDF · Open in the Atlas

Code that ran

21 of 32 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 20 have no contract verdict; 1 failure of our instrument; 11 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/1902.04502.json · badge: /api/ran/1902.04502.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
dbaofd/solar-panels-detection community (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
minus31/Portrait-segmentation community (classifier label)
classifier basis not recorded on the link
8 ran · 9 no recorded run
Tramac/Fast-SCNN-pytorch community (classifier label)
classifier basis not recorded on the link
4 ran · 1 no recorded run
FunctionStatusWhere it lives
batch_intersection_union Ran
ran on a synthesized input
Tramac/Fast-SCNN-pytorch/utils/metric.py
code served (permissive licence) · get_code("b0028c3b51c18eda")
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 2, dtype family int; 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 numpy as np

# Seed for reproducibility
np.random.seed(0)

# Create small prediction and target arrays
# These should be 2D arrays (height x width) with class indices 0 to nclass-1
# Using shape (4, 3) for distinct dimensions
predict = np.array([
    [0, 1, 2],
    [1, 0, 1],
    [2, 2, 0],
    [1, 1, 2]
], dtype=np.int32)

target = np.array([
    [0, 1, 1],
    [1, 0, 2],
    [2, 2, 0],
    [1, 0, 2]
], dtype=np.int32)

# Number of classes (should be > max value in predict/target)
nclass = 3

ARGS = {
    'predict': predict,
    'target': target,
    'nclass': nclass
}
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 tuple, ndim 2, dtype family int — 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
batch_pix_accuracy Ran
ran on a synthesized input
Tramac/Fast-SCNN-pytorch/utils/metric.py
code served (permissive licence) · get_code("d98304f5d56cebc0")
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 int; 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 numpy as np
import torch

torch.manual_seed(0)

# Create small numpy arrays for predict and target
# The function expects 2D arrays (or arrays that can be compared element-wise)
# Let's create arrays of shape (2, 4) with values in range [-1, 1]
# This ensures after +1 conversion they'll be in range [0, 2]
predict = np.array([[-1, 0, 1, -1],
                    [0, 1, -1, 0]], dtype=np.float32)

target = np.array([[-1, 0, 1, 0],
                   [0, 1, -1, 1]], dtype=np.float32)

ARGS = {
    'predict': predict,
    'target': target
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — 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
bilinear_interpolation Ran
ran on a synthesized input
minus31/Portrait-segmentation/fast_scnn.py
code served (permissive licence) · get_code("fe1f1ddb18f26347")
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, 6, 5, 2) 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 tensor
# Shape: (batch_size, height, width, channels)
# Using distinct dimensions: batch=2, height=4, width=3, channels=2
x = tf.constant(np.random.randn(2, 4, 3, 2).astype(np.float32))

# Target size for resizing: (new_height, new_width)
size = [6, 5]

ARGS = {
    "x": x,
    "size": size,
}
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, 6, 5, 2) — shape (2, 6, 5, 2) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
bottleneck_block Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/fast_scnn_0.py
code served (permissive licence) · get_code("5694a1e205acecda")
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; 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 tensorflow as tf

# Create a small input tensor for bottleneck_block
# Shape: (batch_size=2, height=4, width=3, channels=16)
inputs = tf.random.normal((2, 4, 3, 16))

# Define arguments for bottleneck_block
ARGS = {
    'inputs': inputs,
    'filters': 32,           # output filters
    'kernel': (3, 3),        # depthwise convolution kernel size
    't': 6,                  # expansion factor (multiplier for intermediate channels)
    'strides': 1,            # stride for first block
    'n': 2                   # number of residual blocks
}
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 EagerTensor, ndim 4, dtype family float — no output shape is stated: observe_signature.py could not observe it (import:TypeError)
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
composite4 Ran
ran on a synthesized input
minus31/Portrait-segmentation/indexnet/Composition_code.py
code served (permissive licence) · get_code("e41c7308aeed9225")
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 (5, 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 numpy as np

# Create small test images
# fg: foreground image (h, w, 3) or similar - needs to be indexable like an array
fg = np.random.randint(0, 256, size=(5, 4, 3), dtype=np.uint8)

# bg: background image (h, w, 3) - needs to support slicing [0:h, 0:w]
bg = np.random.randint(0, 256, size=(10, 8, 3), dtype=np.uint8)

# a: alpha values (h, w) - will be divided by 255.0
a = np.random.randint(0, 256, size=(5, 4), dtype=np.uint8)

# w, h: width and height dimensions
w = 4
h = 5

ARGS = {
    "fg": fg,
    "bg": bg,
    "a": a,
    "w": w,
    "h": h
}
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 ndarray, ndim 3, dtype family int, shape (5, 4, 3) — shape (5, 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
conv_block Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/fast_scnn_0.py
code served (permissive licence) · get_code("721b77a7a5de2adf")
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 KerasTensor. 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 a small input tensor (batch_size=2, height=4, width=4, channels=3)
inputs = tf.keras.Input(shape=(4, 4, 3))

# Define all parameters for conv_block
ARGS = {
    'inputs': inputs,
    'conv_type': 'ds',  # Using depthwise separable convolution
    'kernel': 16,  # Number of filters
    'kernel_size': 3,  # Kernel size
    'strides': 1,  # Strides
    'padding': 'same',  # Padding mode
    'relu': True  # Apply ReLU activation
}
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 KerasTensor — a KerasTensor 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
conv_block_3 Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/segnet_2.py
code served (permissive licence) · get_code("69ebb6a685a47229")
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 KerasTensor. 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
from tensorflow.keras import layers, backend
import numpy as np

# Create a small input tensor
# Shape: (batch_size, height, width, channels)
# batch=2, height=4, width=4, channels=64
input_tensor = tf.keras.Input(shape=(4, 4, 64))

# Create a symbolic tensor that can be passed through the function
# We need to create an actual tensor for the function to work
batch_size = 2
height = 4
width = 4
input_channels = 64
input_data = tf.random.normal((batch_size, height, width, input_channels))

# Define parameters
kernel_size = 3
filters = [32, 32, 64]  # filters1, filters2, filters3
stage = 2
block = 'a'

ARGS = {
    'input_tensor': input_tensor,
    'kernel_size': kernel_size,
    'filters': filters,
    'stage': stage,
    'block': block
}
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 KerasTensor — a KerasTensor 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
crop_image Ran
ran on a synthesized input
dbaofd/solar-panels-detection/data_processing/data_processing_tool_1.py
code served (permissive licence) · get_code("fa7eb507337ce943")
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 Image. 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 numpy as np

# Create a small test image as a numpy array
# The function expects either uint8 or float values that will be scaled to 0-255
test_image = np.random.rand(512, 512, 3).astype(np.float32)

ARGS = {
    "img": test_image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
crop_image Ran
ran on a synthesized input
dbaofd/solar-panels-detection/data_processing/data_processing_tool_2.py
code served (permissive licence) · get_code("01bf9e1fd4ec484d")
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 4, dtype family int; 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 numpy as np
from PIL import Image

# Create a sample image array
# The function expects either uint8 or float32 format
# Create an image that's a multiple of CROP_WIDTH and CROP_HEIGHT (256x256)
# Let's make a 512x512 image to get multiple crops
sample_image = np.random.randint(0, 256, size=(512, 512, 3), dtype=np.uint8)

ARGS = {
    "img": sample_image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 4, dtype family int — 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
dilated_conv_3x3_bn Ran
ran on a synthesized input
minus31/Portrait-segmentation/indexnet/hlaspp.py
code served (permissive licence) · get_code("b7f80e16cc196e71")
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 Sequential. 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 = {
    'inp': 3,
    'oup': 16,
    'padding': 1,
    'dilation': 1,
    'BatchNorm2d': nn.BatchNorm2d
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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_color_pallete Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Tramac/Fast-SCNN-pytorch/utils/visualize.py
code served (permissive licence) · get_code("1b78db99bae50239")
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 Image. 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 numpy as np
from PIL import Image

# Create a small numpy array for a segmentation map
# Shape: H=4, W=5, single channel
npimg = np.random.randint(0, 19, size=(4, 5), dtype=np.int32)

ARGS = {
    'npimg': npimg,
    'dataset': 'citys'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 2 and dtype family int 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
hflip Ran
ran on a synthesized input
minus31/Portrait-segmentation/augmentation.py
code served (permissive licence) · get_code("6c6cdac56400f7eb")
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 3, 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 numpy as np

# Set seed for reproducibility
np.random.seed(0)

# Create small test images
img = np.random.rand(4, 3).astype(np.float32)  # 4x3 image
mask = np.random.randint(0, 2, (4, 3)).astype(np.float32)  # 4x3 mask

ARGS = {
    'img': img,
    'mask': mask
}
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 tuple, ndim 3, 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
identity_block Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/segnet_3.py
code served (permissive licence) · get_code("f73ed37393ddb228")
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 KerasTensor. 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

# Set random seed for reproducibility
tf.random.set_seed(0)
np.random.seed(0)

# Create a small input tensor with shape (batch, height, width, channels)
# Using channels_last format (default for TensorFlow)
input_tensor = tf.keras.Input(shape=(4, 4, 64))

# Define parameters
kernel_size = 3
filters = [64, 64, 64]  # List of 3 integers for the 3 conv layers
stage = 1
block = 'a'
next_layer_filter = 64

ARGS = {
    'input_tensor': input_tensor,
    'kernel_size': kernel_size,
    'filters': filters,
    'stage': stage,
    'block': block,
    'next_layer_filter': next_layer_filter
}
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 KerasTensor — a KerasTensor 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
identity_block_3 Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/segnet_2.py
code served (permissive licence) · get_code("45aae890a696aa1d")
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 KerasTensor. 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
from tensorflow.keras import layers, backend
import numpy as np

# Set random seeds for reproducibility
np.random.seed(0)
tf.random.set_seed(0)

# Create a small input tensor
# Shape: (batch_size, height, width, channels)
# Using batch=2, height=4, width=3, channels=64
input_tensor = tf.keras.Input(shape=(4, 3, 64), batch_size=2)

# Define parameters
kernel_size = 3
filters = [32, 32, 64]  # filters1, filters2, filters3
stage = 1
block = 'a'
next_layer_filter = 64

ARGS = {
    'input_tensor': input_tensor,
    'kernel_size': kernel_size,
    'filters': filters,
    'stage': stage,
    'block': block,
    'next_layer_filter': next_layer_filter
}
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 KerasTensor — a KerasTensor 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
identity_img Ran
ran on a synthesized input
minus31/Portrait-segmentation/augmentation.py
code served (permissive licence) · get_code("b88010e1312d8957")
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 torch
import numpy as np

torch.manual_seed(0)

# The function identity_img takes img and mask as parameters
# and returns them unchanged. It accepts *args and **kwargs but doesn't use them.
# We need to provide valid image and mask tensors.

# Create a small image tensor (batch=2, channels=3, height=4, width=5)
img = torch.randn(2, 3, 4, 5)

# Create a small mask tensor (batch=2, height=4, width=5)
mask = torch.randint(0, 2, (2, 4, 5))

ARGS = {
    'img': img,
    'mask': mask,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — 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
load_img_mask_pair Ran
ran on a synthesized input
minus31/Portrait-segmentation/get_matte_mask_from_binary_mask.py
code served (permissive licence) · get_code("ffc64df0f93c1c55")
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 numpy as np
import cv2
import tempfile
import os

# Create a temporary directory structure to simulate the expected paths
temp_dir = tempfile.mkdtemp()

# Create subdirectories
img_dir = os.path.join(temp_dir, 'img')
mask_dir = os.path.join(temp_dir, 'mask')
os.makedirs(img_dir, exist_ok=True)
os.makedirs(mask_dir, exist_ok=True)

# Create a small test image (100x100 RGB)
test_img = np.random.randint(0, 255, (100, 100, 3), dtype=np.uint8)
img_path = os.path.join(img_dir, 'test_image.jpeg')
cv2.imwrite(img_path, test_img)

# Create corresponding mask
test_mask = np.random.randint(0, 2, (100, 100), dtype=np.uint8)
mask_path = os.path.join(mask_dir, 'test_image.png')
cv2.imwrite(mask_path, test_mask)

ARGS = {
    'img_path': img_path
}
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 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
make_trimap Ran
ran on a synthesized input
minus31/Portrait-segmentation/get_matte_mask_from_binary_mask.py
code served (permissive licence) · get_code("ebff3c686211b2ae")
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 2, dtype family float, shape (64, 64) 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
import cv2

# Create a simple binary mask for testing
# The function expects a 2D array with values that can be 0-255 or 0-1
mask = np.random.randint(0, 2, size=(64, 64)).astype(np.uint8) * 255

# Ensure the mask has some foreground and background
mask[10:20, 10:20] = 255  # foreground region
mask[40:50, 40:50] = 0    # background region

ARGS = {
    "mask": mask,
    "size": (10, 10)
}
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 2, dtype family float, shape (64, 64) — shape (64, 64) 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
pixelAccuracy Ran
ran on a synthesized input
Tramac/Fast-SCNN-pytorch/utils/metric.py
code served (permissive licence) · get_code("5a4af1d12a8f04f0")
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 numpy as np

# Create simple numpy arrays for predictions and labels
# imPred: predicted class labels for each pixel
# imLab: ground truth class labels for each pixel
# Using small spatial dimensions (4x3) for distinct sizes

imPred = np.array([
    [0, 1, 2, 1],
    [1, 1, 0, 2],
    [2, 0, 1, 0]
], dtype=np.int64)

imLab = np.array([
    [0, 1, 2, 1],
    [1, 0, 0, 2],
    [2, 0, 1, -1]  # -1 indicates unlabeled/invalid pixel
], dtype=np.int64)

ARGS = {
    'imPred': imPred,
    'imLab': imLab
}
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 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
pyramid_pooling_block Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/fast_scnn_0.py
code served (permissive licence) · get_code("f61ffd42fa393081")
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, 8, 387) 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

# Create a small input tensor with shape (batch_size, height, width, channels)
# batch=2, height=8, width=8, channels=3
input_tensor = tf.random.normal((2, 8, 8, 3))

# Define bin_sizes - a list of integers representing pyramid levels
bin_sizes = [1, 2, 4]

ARGS = {
    'input_tensor': input_tensor,
    'bin_sizes': bin_sizes,
}
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 EagerTensor, ndim 4, dtype family float, shape (2, 8, 8, 387) — shape (2, 8, 8, 387) 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
pyramid_pooling_block Ran
ran on a synthesized input
dbaofd/solar-panels-detection/model_list/fast_scnn_1.py
code served (permissive licence) · get_code("a2fc73a1ddb8d26a")
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; 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 tensorflow as tf
import numpy as np

# Create a small input tensor with shape (batch=2, height=32, width=32, channels=3)
input_tensor = tf.constant(
    np.random.randn(2, 32, 32, 3).astype(np.float32)
)

# Define bin sizes for pyramid pooling
bin_sizes = [1, 2, 4]

ARGS = {
    'input_tensor': input_tensor,
    'bin_sizes': bin_sizes
}
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 — no output shape is stated: observe_signature.py recorded SymbolicTensor ndim 4, which disagrees with the fixture run
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
random_activate Ran
ran on a synthesized input
minus31/Portrait-segmentation/augmentation.py
code served (permissive licence) · get_code("424755265e383fb2")
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 function. 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 numpy as np
import torch

def identity_img(img, mask, *args, **kwargs):
    return (img, mask)

def random_activate(aug_func):
    def wrapper(*args, **kwargs):
        if np.random.choice([0, 1]):
            return aug_func(*args, **kwargs)
        else:
            return identity_img(*args, **kwargs)
    return wrapper

# Create a simple augmentation function to pass to random_activate
def simple_aug_func(img, mask, *args, **kwargs):
    # Apply some simple transformation
    return (img * 0.9, mask * 0.9)

ARGS = {
    "aug_func": simple_aug_func
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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_block Not yet run
unverified (harvested, no recorded run)
dbaofd/solar-panels-detection/model_list/segnet_3.py
code served (permissive licence) · get_code("740cb4ae8f653627")
How it ran

no invocation recorded.

depth_sep_dilated_conv_3x3_bn Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/indexnet/hlaspp.py
code served (permissive licence) · get_code("145fc345406f0e26")
How it ran

no invocation recorded.

get_edge Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/data_generator.py
code served (permissive licence) · get_code("c35601c93ff05d52")
How it ran

no invocation recorded.

get_fast_scnn Not yet run
unverified (harvested, no recorded run)
Tramac/Fast-SCNN-pytorch/models/fast_scnn.py
code served (permissive licence) · get_code("acb7752b94575ac5")
How it ran

no invocation recorded.

linearBottleneck Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/fast_scnn.py
code served (permissive licence) · get_code("8686920e76a2f219")
How it ran

no invocation recorded.

network_big Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/network.py
code served (permissive licence) · get_code("1924371588f8b45b")
How it ran

no invocation recorded.

network_small Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/network.py
code served (permissive licence) · get_code("2f1f371d25d86691")
How it ran

no invocation recorded.

read_image Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/get_matte_mask_from_binary_mask.py
code served (permissive licence) · get_code("5eba125ca4ddd12a")
How it ran

no invocation recorded.

residual_block Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/network.py
code served (permissive licence) · get_code("40bf1a25c403d94a")
How it ran

no invocation recorded.

rotate90_img Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/data_generator.py
code served (permissive licence) · get_code("71303057584fee4a")
How it ran

no invocation recorded.

swish_activation Not yet run
unverified (harvested, no recorded run)
minus31/Portrait-segmentation/fast_scnn.py
code served (permissive licence) · get_code("29e46b6d9898edb6")
How it ran

no invocation recorded.

Also present in other papers' code

Of 32 samples lifted from this paper's linked repositories, 32 carry a function identity and 0 do not. Of the 32 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 1 later) and 30 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
composite4 1 other paper
1 dated earlier
1703.03872 (2017-03)
Deep Image Matting
get_color_pallete 1 other paper
1 later
1911.09070 (2019-11)
EfficientDet: Scalable and Efficient Object Detection

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The encoder-decoder framework is state-of-the-art for offline semantic image segmentation. Since the rise in autonomous systems, real-time computation is increasingly desirable. In this paper, we introduce fast segmentation convolutional neural network (Fast-SCNN), an above real-time semantic segmentation model on high resolution image data (1024x2048px) suited to efficient computation on embedded devices with low memory. Building on existing two-branch methods for fast segmentation, we introduce our `learning to downsample' module which computes low-level features for multiple resolution branches simultaneously. Our network combines spatial detail at high resolution with deep features extracted at lower resolution, yielding an accuracy of 68.0% mean intersection over union at 123.5 frames per second on Cityscapes. We also show that large scale pre-training is unnecessary. We thoroughly validate our metric in experiments with ImageNet pre-training and the coarse labeled data of Cityscapes. Finally, we show even faster computation with competitive results on subsampled inputs, without any network modifications.

For agents

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

get_harvested_code_for_paper("1902.04502")
get_code_for_paper("1902.04502")
have("1902.04502")

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

curl https://syntology.ai/api/ran/1902.04502.json

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

[![Syntology run record](https://syntology.ai/api/ran/1902.04502.svg)](https://syntology.ai/paper/1902.04502)

Connect an agent — have() is free.