SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1908.09492 · 2019

Class-balanced Grouping and Sampling for Point Cloud 3D Object Detection

arXiv · PDF · Open in the Atlas

Code that ran

20 of 22 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; 1 honoured the contract we drafted, 0 violated it and 13 have no contract verdict; 6 failures of our instrument; 2 no recorded run), measured 2026-11-21; 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/1908.09492.json · badge: /api/ran/1908.09492.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
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection canonical (classifier label)
classifier basis: census
9 ran · 1 no recorded run
poodarchu/Det3D community (classifier label)
classifier basis not recorded on the link
11 ran · 1 no recorded run
FunctionStatusWhere it lives
bbox_overlaps Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
poodarchu/Det3D/det3d/core/bbox/iou.py
code served (permissive licence) · get_code("93aa32a1463f1fad")
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, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create two sets of bounding boxes in format (x1, y1, x2, y2)
# bboxes1: shape (m, 4) where m=2
bboxes1 = torch.tensor([
    [10.0, 20.0, 30.0, 40.0],
    [15.0, 25.0, 35.0, 45.0]
], dtype=torch.float32)

# bboxes2: shape (n, 4) where n=3
bboxes2 = torch.tensor([
    [12.0, 22.0, 32.0, 42.0],
    [20.0, 30.0, 40.0, 50.0],
    [5.0, 5.0, 15.0, 15.0]
], dtype=torch.float32)

# For is_aligned=True case, we need equal number of boxes
bboxes1_aligned = torch.tensor([
    [10.0, 20.0, 30.0, 40.0],
    [15.0, 25.0, 35.0, 45.0]
], dtype=torch.float32)

bboxes2_aligned = torch.tensor([
    [12.0, 22.0, 32.0, 42.0],
    [20.0, 30.0, 40.0, 50.0]
], dtype=torch.float32)

ARGS = {
    "bboxes1": bboxes1,
    "bboxes2": bboxes2,
    "mode": "iou",
    "is_aligned": 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 Tensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
bound_points_jit Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/ops/point_cloud/point_cloud_ops.py
code served (permissive licence) · get_code("e5993ab3f1529dc7")
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 1, dtype family int, shape (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 numba
import numpy as np

# Create small sample data
points = np.array([
    [0.5, 1.5, 2.5, 3.5],
    [1.0, 2.0, 3.0, 4.0],
], dtype=np.float64)

upper_bound = np.array([2.0, 3.0, 4.0, 5.0], dtype=np.float64)
lower_bound = np.array([0.0, 1.0, 2.0, 3.0], dtype=np.float64)

ARGS = {
    "points": points,
    "upper_bound": upper_bound,
    "lower_bound": lower_bound,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (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
create_target_np Ran
ran; our contract draft was wrong, not the code
poodarchu/Det3D/det3d/core/anchor/target_ops.py
code served (permissive licence) · get_code("20f05c72a0b0b39d")
How it ran

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

import numpy as np
import numpy.random as npr

# Set seed for reproducibility
npr.seed(0)

# Define small helper functions for similarity and box encoding
def similarity_fn(anchors, gt_boxes):
    """Compute IoU-like similarity matrix between anchors and gt_boxes"""
    # Return a matrix of shape (num_anchors, num_gt_boxes)
    num_anchors = anchors.shape[0]
    num_gt_boxes = gt_boxes.shape[0]
    # Simple mock: random similarity scores between 0 and 1
    return np.random.rand(num_anchors, num_gt_boxes)

def box_encoding_fn(gt_boxes, anchors):
    """Encode gt_boxes relative to anchors"""
    # Return box encodings of same shape as anchors
    return np.random.randn(*anchors.shape)

def prune_anchor_fn(all_anchors):
    """Keep only valid anchors (mock: keep first 80% of them)"""
    num_anchors = all_anchors.shape[0]
    num_valid = max(1, int(0.8 * num_anchors))
    return np.arange(num_valid)

# Create small test data
num_anchors = 10
num_gt_boxes = 3
box_ndim = 7

all_anchors = np.random.randn(num_anchors, box_ndim).astype(np.float32)
gt_boxes = np.random.randn(num_gt_boxes, box_ndim).astype(np.float32)
gt_classes = np.array([1, 2, 1], dtype=np.int32)

ARGS = {
    'all_anchors': all_anchors,
    'gt_boxes': gt_boxes,
    'similarity_fn': similarity_fn,
    'box_encoding_fn': box_encoding_fn,
    'prune_anchor_fn': prune_anchor_fn,
    'gt_classes': gt_classes,
    'matched_threshold': 0.6,
    'unmatched_threshold': 0.45,
    'bbox_inside_weight': None,
    'positive_fraction': 0.5,
    'rpn_batch_size': 300,
    'norm_by_num_examples': False,
    'box_code_size': 7,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
div_up Ran
ran, honoured the contract we drafted
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/kitti_object_eval_python/rotate_iou.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from datongjie/bevspread, is pointer only (licence: NONE) · get_code("20fbfa2279e1ce4b")
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 datongjie/bevspread, 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.

get_image_index_str Ran
ran; our contract draft was wrong, not the code
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/kitti_object_eval_python/kitti_common.py
code served (permissive licence) · get_code("3415a0d1c950c1c7")
How it ran

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

import torch

torch.manual_seed(0)

ARGS = {
    "img_idx": 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_image_path Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/kitti_object_eval_python/kitti_common.py
code served (permissive licence) · get_code("c8b847449b3be5dc")
How it ran

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

import pathlib
import tempfile
import os

# Create a temporary directory structure to satisfy the file existence check
temp_dir = tempfile.mkdtemp()
training_dir = pathlib.Path(temp_dir) / 'training' / 'image_2'
training_dir.mkdir(parents=True, exist_ok=True)

# Create a dummy image file
dummy_image = training_dir / '000000.png'
dummy_image.touch()

# Define arguments for get_image_path
ARGS = {
    'idx': 0,
    'prefix': temp_dir,
    'training': True,
    'relative_path': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
inverse_sigma Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/ground_plane_detection.py
code served (permissive licence) · get_code("363a2d53034b488a")
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 MatrixLike. 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)

class MatrixLike:
    """Stand-in class that mimics matrix behavior with copy() and transpose() methods"""
    def __init__(self, data):
        self.data = data
    
    def copy(self):
        """Return a copy of self with copied data"""
        new_obj = MatrixLike(self.data.copy())
        return new_obj
    
    def transpose(self):
        """Return transposed version"""
        transposed = MatrixLike(self.data.T)
        return transposed
    
    def __getitem__(self, key):
        return self.data[key]
    
    def __setitem__(self, key, value):
        self.data[key] = value
    
    @property
    def shape(self):
        return self.data.shape

# Create a small matrix for testing
# Use a 4x3 matrix to have distinct dimensions
matrix_data = np.array([
    [2.0, 0.5, 0.1],
    [0.0, 4.0, 0.2],
    [0.0, 0.0, 0.5],
    [0.1, 0.2, 0.3]
], dtype=np.float32)

S = MatrixLike(matrix_data)

ARGS = {
    "S": S
}
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 MatrixLike — a MatrixLike output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MatrixLike 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
is_line_segment_intersection_jit Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
poodarchu/Det3D/det3d/core/bbox/geometry.py
code served (permissive licence) · get_code("0513b9d43a6d7883")
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 (2, 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
import numba

# Create line segments for lines1: shape [N, 2, 2]
# N=2 line segments, each with 2 endpoints, each endpoint has 2 coordinates (x, y)
lines1 = np.array([
    [[0.0, 0.0], [1.0, 1.0]],  # Line segment 1
    [[0.0, 1.0], [1.0, 0.0]]   # Line segment 2
], dtype=np.float64)

# Create line segments for lines2: shape [M, 2, 2]
# M=3 line segments
lines2 = np.array([
    [[0.5, 0.0], [0.5, 1.0]],  # Line segment 1
    [[0.0, 0.5], [1.0, 0.5]],  # Line segment 2
    [[2.0, 2.0], [3.0, 3.0]]   # Line segment 3
], dtype=np.float64)

ARGS = {
    "lines1": lines1,
    "lines2": lines2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, 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 (2, 3) — shape (2, 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
line_segment_intersection Ran
ran on a synthesized input
poodarchu/Det3D/det3d/core/bbox/geometry.py
code served (permissive licence) · get_code("6a7ad14559c54076")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import numba

# Create numpy arrays for the line segments
# line1 represents a line segment from point A to point B
# line2 represents a line segment from point C to point D
line1 = np.array([[0.0, 0.0], [2.0, 2.0]], dtype=np.float64)
line2 = np.array([[0.0, 2.0], [2.0, 0.0]], dtype=np.float64)

# Create an output array to store the intersection point
intersection = np.array([0.0, 0.0], dtype=np.float64)

ARGS = {
    "line1": line1,
    "line2": line2,
    "intersection": intersection
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
make_dot Ran
ran on a synthesized input
poodarchu/Det3D/det3d/visualization/netviz.py
code served (permissive licence) · get_code("78f63746495b109f")
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 Digraph. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch.autograd import Variable

torch.manual_seed(0)

# Create a simple computation graph with a variable that requires gradients
x = torch.randn(2, 4, requires_grad=True)
w = torch.randn(4, 3, requires_grad=True)

# Perform operations to build autograd graph
y = torch.matmul(x, w)
z = y.sum()

# Create params dict for labeling
params = {
    'x': x,
    'w': w
}

ARGS = {
    'var': z,
    'params': params
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
graphviz, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Digraph — a Digraph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Digraph 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
points_count_convex_polygon_3d_jit Ran
ran on a synthesized input
poodarchu/Det3D/det3d/core/bbox/geometry.py
code served (permissive licence) · get_code("0dba4924dbd3544f")
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 1, dtype family int, shape (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 numpy as np

# Create small test data for the function
# points: [num_points, 3] array
num_points = 2
points = np.array([
    [0.5, 0.5, 0.5],
    [1.5, 1.5, 1.5]
], dtype=np.float64)

# polygon_surfaces: [num_polygon, max_num_surfaces, max_num_points_of_surface, 3]
# Create a simple convex polygon (a tetrahedron-like structure)
num_polygons = 2
max_num_surfaces = 4
max_num_points_of_surface = 3

polygon_surfaces = np.zeros(
    (num_polygons, max_num_surfaces, max_num_points_of_surface, 3),
    dtype=np.float64
)

# Define simple triangular surfaces for first polygon
# Triangle 1 of polygon 0
polygon_surfaces[0, 0, 0] = [0.0, 0.0, 0.0]
polygon_surfaces[0, 0, 1] = [1.0, 0.0, 0.0]
polygon_surfaces[0, 0, 2] = [0.0, 1.0, 0.0]

# Triangle 2 of polygon 0
polygon_surfaces[0, 1, 0] = [0.0, 0.0, 0.0]
polygon_surfaces[0, 1, 1] = [0.0, 1.0, 0.0]
polygon_surfaces[0, 1, 2] = [0.0, 0.0, 1.0]

# Triangle 3 of polygon 0
polygon_surfaces[0, 2, 0] = [0.0, 0.0, 0.0]
polygon_surfaces[0, 2, 1] = [0.0, 0.0, 1.0]
polygon_surfaces[0, 2, 2] = [1.0, 0.0, 0.0]

# Triangle 4 of polygon 0
polygon_surfaces[0, 3, 0] = [1.0, 0.0, 0.0]
polygon_surfaces[0, 3, 1] = [0.0, 1.0, 0.0]
polygon_surfaces[0, 3, 2] = [0.0, 0.0, 1.0]

# Define simple triangular surfaces for second polygon (offset)
polygon_surfaces[1, 0, 0] = [1.0, 1.0, 1.0]
polygon_surfaces[1, 0, 1] = [2.0, 1.0, 1.0]
polygon_surfaces[1, 0, 2] = [1.0, 2.0, 1.0]

polygon_surfaces[1, 1, 0] = [1.0, 1.0, 1.0]
polygon_surfaces[1, 1, 1] = [1.0, 2.0, 1.0]
polygon_surfaces[1, 1, 2] = [1.0, 1.0, 2.0]

polygon_surfaces[1, 2, 0] = [1.0, 1.0, 1.0]
polygon_surfaces[1, 2, 1] = [1.0, 1.0, 2.0]
polygon_surfaces[1, 2, 2] = [2.0, 1.0, 1.0]

polygon_surfaces[1, 3, 0] = [2.0, 1.0, 1.0]
polygon_surfaces[1, 3, 1] = [1.0, 2.0, 1.0]
polygon_surfaces[1, 3, 2] = [1.0, 1.0, 2.0]

# num_surfaces: [num_polygon] array indicating how many surfaces each polygon has
num_surfaces = np.array([4, 4], dtype=np.int64)

ARGS = {
    "points": points,
    "polygon_surfaces": polygon_surfaces,
    "num_surfaces": num_surfaces
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (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
points_to_bev Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/ops/point_cloud/bev_ops.py
code served (permissive licence) · get_code("0a6df8e3e11be41b")
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 float, shape (3, 20, 20) 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 sample point cloud data
# Shape: [N, 4] where columns are x, y, z, reflectivity
points = np.array([
    [0.5, 0.5, 0.5, 0.8],
    [1.2, 1.3, 0.6, 0.9],
], dtype=np.float32)

# Voxel size for x, y, z dimensions
voxel_size = [0.1, 0.1, 0.8]

# Coordinate range: [x_min, y_min, z_min, x_max, y_max, z_max]
coors_range = [0.0, 0.0, 0.0, 2.0, 2.0, 2.0]

ARGS = {
    "points": points,
    "voxel_size": voxel_size,
    "coors_range": coors_range,
    "with_reflectivity": False,
    "density_norm_num": 16,
    "max_voxels": 40000,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, 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 float, shape (3, 20, 20) — shape (3, 20, 20) 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
points_to_voxel Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/ops/point_cloud/point_cloud_ops.py
code served (permissive licence) · get_code("7ce4e5103722da22")
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

# Create small sample point cloud data
# N=10 points, each with 4 dimensions (x, y, z, reflectivity)
np.random.seed(0)
points = np.random.randn(10, 4).astype(np.float32)
# Scale points to be within a reasonable range
points = points * 10 + 50

# Define voxel size [x, y, z]
voxel_size = np.array([0.1, 0.1, 0.1], dtype=np.float32)

# Define coordinate range [x_min, y_min, z_min, x_max, y_max, z_max]
coors_range = np.array([40.0, 40.0, 40.0, 60.0, 60.0, 60.0], dtype=np.float32)

# Maximum points per voxel
max_points = 35

# Whether to reverse the index (zyx instead of xyz)
reverse_index = True

# Maximum number of voxels
max_voxels = 20000

ARGS = {
    "points": points,
    "voxel_size": voxel_size,
    "coors_range": coors_range,
    "max_points": max_points,
    "reverse_index": reverse_index,
    "max_voxels": max_voxels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, 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
replace Ran
ran on a synthesized input
poodarchu/Det3D/det3d/visualization/netviz.py
code served (permissive licence) · get_code("34a445a4ab172be6")
How it ran

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

import torch

torch.manual_seed(0)

ARGS = {
    'name': 'example_key',
    'scope': {'example_key': 'prefix_value'}
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
second_box_decode Ran
ran on a synthesized input
poodarchu/Det3D/det3d/core/bbox/box_torch_ops.py
code served (permissive licence) · get_code("755dd896938eee95")
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, 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 torch

torch.manual_seed(0)

# Test case 1: box_ndim=7, encode_angle_to_vector=False, smooth_dim=False, norm_velo=False
box_encodings_7 = torch.randn(2, 7)
anchors_7 = torch.randn(2, 7)

# Test case 2: box_ndim=9, encode_angle_to_vector=True, smooth_dim=True, norm_velo=True
box_encodings_9_vec = torch.randn(2, 10)
anchors_9 = torch.randn(2, 9)

# Test case 3: box_ndim=9, encode_angle_to_vector=False, smooth_dim=False, norm_velo=False
box_encodings_9 = torch.randn(2, 9)

ARGS = {
    "box_encodings": box_encodings_7,
    "anchors": anchors_7,
    "encode_angle_to_vector": False,
    "bin_loss": False,
    "smooth_dim": False,
    "norm_velo": 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 Tensor, ndim 2, dtype family float, shape (2, 7) — shape (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
second_box_encode Ran
ran on a synthesized input
poodarchu/Det3D/det3d/core/bbox/box_torch_ops.py
code served (permissive licence) · get_code("cba7c7fe7a6c4de3")
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, 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 torch

torch.manual_seed(0)

# Create small tensors for boxes and anchors
# boxes: [N, 7] with x, y, z, l, w, h, r
boxes = torch.randn(2, 7)

# anchors: [N, 7] with x, y, z, l, w, h, r
anchors = torch.randn(2, 7)

ARGS = {
    "boxes": boxes,
    "anchors": anchors,
    "encode_angle_to_vector": False,
    "smooth_dim": False,
    "norm_velo": 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 Tensor, ndim 2, dtype family float, shape (2, 7) — shape (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
svd Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/ground_plane_detection.py
code served (permissive licence) · get_code("0215535ccf00014e")
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

# The function svd(A) takes a matrix A and computes its SVD decomposition
# It uses numpy.linalg.svd and constructs a diagonal matrix S with the singular values
# A should be a numpy array with shape (m, n) where m and n are different for clarity

# Create a small test matrix with distinct dimensions
A = np.array([
    [1.0, 2.0, 3.0],
    [4.0, 5.0, 6.0],
    [7.0, 8.0, 9.0],
    [10.0, 11.0, 12.0]
], dtype=np.float64)

ARGS = {
    "A": A
}
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 — 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
svd_solve Ran
ran on a synthesized input
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/ground_plane_detection.py
code served (permissive licence) · get_code("d9f32b1e24612346")
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 1, dtype family float, shape (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 a simple system: A @ x = b
# Using a 4x3 matrix A and a 4-element vector b
# This creates an overdetermined system (more equations than unknowns)

np.random.seed(0)

# Create A as a 4x3 matrix
A = np.random.randn(4, 3).astype(np.float64)

# Create b as a 4-element vector
b = np.random.randn(4).astype(np.float64)

ARGS = {
    'A': A,
    'b': b
}
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 1, dtype family float, shape (3,) — shape (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
torch_to_np_dtype Ran
ran on a synthesized input
poodarchu/Det3D/det3d/core/bbox/box_torch_ops.py
code served (permissive licence) · get_code("4dcdb0a71119c780")
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 Float32DType. 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

# Call torch.manual_seed for reproducibility
torch.manual_seed(0)

# The function torch_to_np_dtype takes a single parameter: ttype
# which is a torch dtype, and returns the corresponding numpy dtype
# by looking it up in a dictionary.

# Valid torch dtypes to test with
ttype = torch.float32

# Construct the ARGS dictionary
ARGS = {
    'ttype': ttype
}
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 Float32DType — a Float32DType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Float32DType 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
unmap Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
poodarchu/Det3D/det3d/core/anchor/target_ops.py
code served (permissive licence) · get_code("2902bedb7c18b1a1")
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 1, dtype family int, shape (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 numpy as np

# Test case 1: 1D array with subset indices
data_1d = np.array([10, 20, 30], dtype=np.int32)
count_1d = 5
inds_1d = np.array([1, 3, 4], dtype=np.int64)
fill_1d = 0

# Test case 2: 2D array with subset indices
data_2d = np.array([[1, 2], [3, 4], [5, 6]], dtype=np.float32)
count_2d = 4
inds_2d = np.array([0, 2, 3], dtype=np.int64)
fill_2d = 0

# Use case 1 for ARGS
ARGS = {
    "data": data_1d,
    "count": count_1d,
    "inds": inds_1d,
    "fill": fill_1d
}
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 1, dtype family int, shape (5,) — shape (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
get_kitti_info_path Not yet run
unverified (harvested, no recorded run)
poodarchu/Class-balanced-Grouping-and-Sampling-for-Point-Cloud-3D-Object-Detection/det3d/datasets/utils/kitti_object_eval_python/kitti_common.py
code served (permissive licence) · get_code("f027b5ae160547ce")
How it ran

no invocation recorded.

parse Not yet run
unverified (harvested, no recorded run)
poodarchu/Det3D/det3d/visualization/netviz.py
code served (permissive licence) · get_code("50ded6a49b80467e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 22 samples lifted from this paper's linked repositories, 22 carry a function identity and 0 do not. Of the 22 distinct functions with an identity, 17 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 17 later) and 5 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
div_up 9 other papers
9 later
1911.10194 (2019-11)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
1911.11288 (2019-11)
Autolabeling 3D Objects with Differentiable Rendering of SDF Shape Priors
1912.13192 (2019-12)
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
and 6 more not named here
get_image_index_str 5 other papers
5 later
1911.10194 (2019-11)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
1912.13192 (2019-12)
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
and 2 more not named here
get_kitti_info_path 4 other papers
1 dated earlier, 3 later
1907.06038 (2019-07)
M3D-RPN: Monocular 3D Region Proposal Network for Object Detection
1911.10194 (2019-11)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
1912.13192 (2019-12)
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
and 1 more not named here
unmap 4 other papers
1 dated earlier, 3 later
1812.05784 (2018-12)
PointPillars: Fast Encoders for Object Detection from Point Clouds
1912.04986 (2019-12)
What You See is What You Get: Exploiting Visibility for 3D Object Detection
2009.00784 (2020-09)
CLOCs: Camera-LiDAR Object Candidates Fusion for 3D Object Detection
and 1 more not named here
get_image_path 3 other papers
3 later
1911.10194 (2019-11)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
1912.13192 (2019-12)
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
is_line_segment_intersection_jit 3 other papers
1 dated earlier, 2 later
1812.05784 (2018-12)
PointPillars: Fast Encoders for Object Detection from Point Clouds
1912.04986 (2019-12)
What You See is What You Get: Exploiting Visibility for 3D Object Detection
2009.00784 (2020-09)
CLOCs: Camera-LiDAR Object Candidates Fusion for 3D Object Detection
line_segment_intersection 3 other papers
1 dated earlier, 2 later
1812.05784 (2018-12)
PointPillars: Fast Encoders for Object Detection from Point Clouds
1912.04986 (2019-12)
What You See is What You Get: Exploiting Visibility for 3D Object Detection
2009.00784 (2020-09)
CLOCs: Camera-LiDAR Object Candidates Fusion for 3D Object Detection
bbox_overlaps 2 other papers
1 dated earlier, 1 later
1906.07155 (2019-06)
MMDetection: Open MMLab Detection Toolbox and Benchmark
2003.13042 (2020-03)
Omni-sourced Webly-supervised Learning for Video Recognition
create_target_np 1 other paper
1 later
2104.09804 (2021-04)
SE-SSD: Self-Ensembling Single-Stage Object Detector From Point Cloud
inverse_sigma 1 other paper
1 later
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
make_dot 1 other paper
1 later
2006.12671 (2020-06)
AFDet: Anchor Free One Stage 3D Object Detection
parse 1 other paper
1 later
2006.12671 (2020-06)
AFDet: Anchor Free One Stage 3D Object Detection
points_to_voxel 1 other paper
1 later
2207.00531 (2022-07)
Masked Autoencoder for Self-Supervised Pre-training on Lidar Point Clouds
replace 1 other paper
1 later
2006.12671 (2020-06)
AFDet: Anchor Free One Stage 3D Object Detection
svd 1 other paper
1 later
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
svd_solve 1 other paper
1 later
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
torch_to_np_dtype 1 other paper
1 later
1912.05163 (2019-12)
TANet: Robust 3D Object Detection from Point Clouds with Triple Attention

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

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

Abstract

This report presents our method which wins the nuScenes3D Detection Challenge [17] held in Workshop on Autonomous Driving(WAD, CVPR 2019). Generally, we utilize sparse 3D convolution to extract rich semantic features, which are then fed into a class-balanced multi-head network to perform 3D object detection. To handle the severe class imbalance problem inherent in the autonomous driving scenarios, we design a class-balanced sampling and augmentation strategy to generate a more balanced data distribution. Furthermore, we propose a balanced group-ing head to boost the performance for the categories withsimilar shapes. Based on the Challenge results, our methodoutperforms the PointPillars [14] baseline by a large mar-gin across all metrics, achieving state-of-the-art detection performance on the nuScenes dataset. Code will be released at CBGS.

For agents

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

get_harvested_code_for_paper("1908.09492")
get_code_for_paper("1908.09492")
have("1908.09492")

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

curl https://syntology.ai/api/ran/1908.09492.json

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

[![Syntology run record](https://syntology.ai/api/ran/1908.09492.svg)](https://syntology.ai/paper/1908.09492)

Connect an agent — have() is free.