SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1412.3128 · 2014

Real-Time Grasp Detection Using Convolutional Neural Networks

arXiv · PDF · Open in the Atlas

Code that ran

10 of 18 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 9 have no contract verdict; 1 failure of our instrument; 8 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/1412.3128.json · badge: /api/ran/1412.3128.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
571502680/robot-grasp-detection community (classifier label)
classifier basis not recorded on the link
6 ran · 5 no recorded run
hamed-hosseini/ggcnn community (classifier label)
classifier basis not recorded on the link
2 ran · 3 no recorded run
DucTranVan/grasp-detection-pytorch community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
bboxes_to_grasps Ran
ran on a synthesized input
571502680/robot-grasp-detection/grasp_det.py
code served (permissive licence) · get_code("ee8e7c782e425ab2")
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 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

# The function takes bboxes as input
# It unstacks along axis=1, expecting 8 elements (box[0] through box[7])
# This means bboxes should have shape (batch_size, 8) or compatible
# The function accesses indices 0,1,2,3,4,5,6,7 after unstacking on axis=1

# Create a small bboxes tensor with batch size 2 and 8 bbox coordinates
bboxes = tf.constant([
    [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0],
    [15.0, 25.0, 35.0, 45.0, 55.0, 65.0, 75.0, 85.0]
], dtype=tf.float32)

ARGS = {
    "bboxes": bboxes
}
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 tuple, ndim 2, 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
bboxes_to_grasps Ran
ran on a synthesized input
DucTranVan/grasp-detection-pytorch/src/models/util.py
code served (permissive licence) · get_code("7fa3dbd955cf872d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 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 torch

torch.manual_seed(0)

# bboxes_to_grasps expects a tensor of bounding boxes
# The function indexes columns 0-7, so we need at least 8 columns
# Each row represents a bounding box in format: [x1, y1, x2, y2, x3, y3, x4, y4]
# where these are typically corner points or similar coordinates

bboxes = torch.tensor([
    [1.0, 2.0, 5.0, 6.0, 8.0, 10.0, 3.0, 4.0],
    [2.0, 3.0, 7.0, 9.0, 11.0, 13.0, 4.0, 5.0]
], dtype=torch.float32)

ARGS = {
    "bboxes": bboxes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 5) — shape (2, 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
conv2d_s1 Ran
ran on a synthesized input
571502680/robot-grasp-detection/grasp_inf.py
code served (permissive licence) · get_code("62d8f8b2389b97f7")
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, 1) 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: batch=2, height=4, width=3, channels=2
x = tf.constant(np.random.randn(2, 4, 3, 2).astype(np.float32))

# Create a small filter tensor: height=3, width=3, in_channels=2, out_channels=1
W = tf.constant(np.random.randn(3, 3, 2, 1).astype(np.float32))

ARGS = {
    'x': x,
    'W': W
}
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, 1) — shape (2, 4, 3, 1) 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
conv2d_s2 Ran
ran on a synthesized input
571502680/robot-grasp-detection/grasp_inf.py
code served (permissive licence) · get_code("3bd95fc813da633d")
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, 2, 2, 7) 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 input tensor x: shape [batch, height, width, channels]
# Using batch=2, height=4, width=3, channels=5 for distinct dimensions
x = tf.constant(np.random.randn(2, 4, 3, 5).astype(np.float32))

# Create weight tensor W: shape [filter_height, filter_width, in_channels, out_channels]
# Using filter_height=3, filter_width=3, in_channels=5, out_channels=7
W = tf.constant(np.random.randn(3, 3, 5, 7).astype(np.float32))

ARGS = {
    'x': x,
    'W': W
}
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, 2, 2, 7) — shape (2, 2, 2, 7) 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
distort_color Ran
ran on a synthesized input
571502680/robot-grasp-detection/grasp_img_proc.py
code served (permissive licence) · get_code("5340de8fabcdb750")
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 (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 tensorflow as tf
import numpy as np

# Create a small image tensor with shape (height, width, channels)
# Using float32 with values in [0, 1] range as expected by tf.image functions
image = tf.constant(np.random.rand(4, 3, 3).astype(np.float32))

# thread_id is an integer
thread_id = 0

ARGS = {
    "image": image,
    "thread_id": thread_id,
}
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 (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
grasp_to_bbox Ran
ran on a synthesized input
571502680/robot-grasp-detection/grasp_det.py
code served (permissive licence) · get_code("d1dd7a0bd6b4800c")
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 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 tensorflow as tf

# Create small TensorFlow tensors for the grasp_to_bbox function
x = tf.constant([100.0, 200.0], dtype=tf.float32)
y = tf.constant([150.0, 250.0], dtype=tf.float32)
tan = tf.constant([0.5, 1.0], dtype=tf.float32)
h = tf.constant([50.0, 60.0], dtype=tf.float32)
w = tf.constant([40.0, 45.0], dtype=tf.float32)

ARGS = {
    "x": x,
    "y": y,
    "tan": tan,
    "h": h,
    "w": w,
}
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 list, ndim 3, 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
grasps_to_bboxes Ran
ran on a synthesized input
DucTranVan/grasp-detection-pytorch/src/models/util.py
code served (permissive licence) · get_code("d4d5a88d7e5cb5d5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 8) 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

torch.manual_seed(0)

# grasps_to_bboxes expects a tensor of shape (batch_size, 5)
# where each row is [x, y, theta, h, w]
grasps = torch.tensor([
    [10.0, 20.0, 0.5, 5.0, 3.0],
    [15.0, 25.0, 1.2, 6.0, 4.0]
], dtype=torch.float32)

ARGS = {
    "grasps": grasps
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 8) — shape (2, 8) 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
max_pool_2x2 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
571502680/robot-grasp-detection/grasp_inf.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from YudeWang/UNet-Satellite-Image-Segmentation, is pointer only (licence: NONE) · get_code("fc52a4f8f0232e03")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from YudeWang/UNet-Satellite-Image-Segmentation, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

post_process_output Ran
ran on a synthesized input
hamed-hosseini/ggcnn/models/common.py
code served (permissive licence) · get_code("1a944ae0e5992dc3")
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 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 torch
from skimage.filters import gaussian

torch.manual_seed(0)

# Create small tensor outputs from a GG-CNN
# The function expects tensors that can be converted to numpy and squeezed
# Typical GG-CNN output shape: (batch, 1, height, width) or (batch, height, width)

batch_size = 2
height = 4
width = 3

# q_img: quality map, shape (batch, 1, height, width)
q_img = torch.randn(batch_size, 1, height, width)

# cos_img and sin_img: used for angle calculation via atan2
# shape (batch, 1, height, width)
cos_img = torch.randn(batch_size, 1, height, width)
sin_img = torch.randn(batch_size, 1, height, width)

# width_img: gripper width prediction, shape (batch, 1, height, width)
# will be multiplied by 150.0
width_img = torch.randn(batch_size, 1, height, width)

ARGS = {
    'q_img': q_img,
    'cos_img': cos_img,
    'sin_img': sin_img,
    'width_img': width_img
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
skimage, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 4, 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
train Ran
ran on a synthesized input
hamed-hosseini/ggcnn/train_cnn.py
code served (permissive licence) · get_code("894f259f11b9d556")
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 torch.nn as nn
import torch.optim as optim

torch.manual_seed(0)

# Define a minimal network that takes input and produces output
class MinimalNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.fc(x)

# Create minimal training data as a list of batches
# Each batch is a tuple of (x, y, idx, rot, zoom_factor)
train_data = [
    (torch.randn(2, 4), torch.randn(2, 3), torch.tensor([0, 1]), torch.tensor([0.0, 0.0]), torch.tensor([1.0, 1.0])),
    (torch.randn(2, 4), torch.randn(2, 3), torch.tensor([2, 3]), torch.tensor([0.0, 0.0]), torch.tensor([1.0, 1.0])),
    (torch.randn(2, 4), torch.randn(2, 3), torch.tensor([4, 5]), torch.tensor([0.0, 0.0]), torch.tensor([1.0, 1.0])),
]

net = MinimalNet()
device = torch.device('cpu')
optimizer = optim.SGD(net.parameters(), lr=0.01)
losses_list = []

ARGS = {
    'epoch': 1,
    'net': net,
    'device': device,
    'train_data': train_data,
    'optimizer': optimizer,
    'batches_per_epoch': 2,
    'losses': losses_list,
    'time': None,
    'vis': False
}
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 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
calculate_iou_match Not yet run
unverified (harvested, no recorded run)
hamed-hosseini/ggcnn/utils/dataset_processing/evaluation.py
code served (permissive licence) · get_code("7eaf73809fb84ee2")
How it ran

no invocation recorded.

calculate_iou_match_hamed Not yet run
unverified (harvested, no recorded run)
hamed-hosseini/ggcnn/utils/dataset_processing/evaluation.py
code served (permissive licence) · get_code("2894e8be5a098961")
How it ran

no invocation recorded.

detect_grasps Not yet run
unverified (harvested, no recorded run)
hamed-hosseini/ggcnn/utils/dataset_processing/grasp.py
code served (permissive licence) · get_code("62c9539b4df460a9")
How it ran

no invocation recorded.

eval_image Not yet run
unverified (harvested, no recorded run)
571502680/robot-grasp-detection/grasp_img_proc.py
code served (permissive licence) · get_code("81b62541aa877dd6")
How it ran

no invocation recorded.

inputs Not yet run
unverified (harvested, no recorded run)
571502680/robot-grasp-detection/archive/grasp_detection.py
code served (permissive licence) · get_code("83abe6bab4413440")
How it ran

no invocation recorded.

parse_example_proto Not yet run
unverified (harvested, no recorded run)
571502680/robot-grasp-detection/grasp_img_proc.py
code served (permissive licence) · get_code("781e15b01fb9e1df")
How it ran

no invocation recorded.

parse_example_proto Not yet run
unverified (harvested, no recorded run)
571502680/robot-grasp-detection/image_processing.py
code served (permissive licence) · get_code("e739a2b8152f12a1")
How it ran

no invocation recorded.

read_and_decode Not yet run
unverified (harvested, no recorded run)
571502680/robot-grasp-detection/archive/grasp_detection.py
code served (permissive licence) · get_code("8265f480063d380c")
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, 3 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 3 later) and 15 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
max_pool_2x2 5 other papers
5 later
1505.04597 (2015-05)
U-Net: Convolutional Networks for Biomedical Image Segmentation
1709.09603 (2017-09)
Riemannian approach to batch normalization
1805.04980 (2018-05)
Unifying and Merging Well-trained Deep Neural Networks for Inference Stage
and 2 more not named here
calculate_iou_match 1 other paper
1 later
1804.05172 (2018-04)
Closing the Loop for Robotic Grasping: A Real-time, Generative Grasp Synthesis Approach
post_process_output 1 other paper
1 later
1804.05172 (2018-04)
Closing the Loop for Robotic Grasping: A Real-time, Generative Grasp Synthesis Approach

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

We present an accurate, real-time approach to robotic grasp detection based on convolutional neural networks. Our network performs single-stage regression to graspable bounding boxes without using standard sliding window or region proposal techniques. The model outperforms state-of-the-art approaches by 14 percentage points and runs at 13 frames per second on a GPU. Our network can simultaneously perform classification so that in a single step it recognizes the object and finds a good grasp rectangle. A modification to this model predicts multiple grasps per object by using a locally constrained prediction mechanism. The locally constrained model performs significantly better, especially on objects that can be grasped in a variety of ways.

For agents

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

get_harvested_code_for_paper("1412.3128")
get_code_for_paper("1412.3128")
have("1412.3128")

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

curl https://syntology.ai/api/ran/1412.3128.json

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

[![Syntology run record](https://syntology.ai/api/ran/1412.3128.svg)](https://syntology.ai/paper/1412.3128)

Connect an agent — have() is free.