SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.13394 · CVPR · 2022

Point2Seq: Detecting 3D Objects as Sequences

Wei Zhang, Xiaogang Wang, Xinchao Wang, Hang Xu, Jiageng Mao, Minzhe Niu, Michael Mi, Yujing Xue

arXiv · PDF · Open in the Atlas

Code that ran

13 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 5 failures of our instrument; 4 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/2203.13394.json · badge: /api/ran/2203.13394.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
ocnflag/point2seq marked official in Papers with Code
Papers with Code archive
11 ran · 4 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 0 no recorded run
FunctionStatusWhere it lives
BoxesIou3dDifferentiableFunction Ran
ran on a synthesized input · constructed an object; no result was computed
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("c3e53caad38b1bd3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, BoxesIou3dDifferentiableFunction: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create mock objects for dependencies that would normally be external
class MockIou3dNmsDiffCuda:
    @staticmethod
    def boxes_overlap_bev_onebyone_gpu(boxes_a, boxes_b, overlaps_bev):
        # Simple mock: compute some overlap values
        # boxes_a and boxes_b are [N, 7], overlaps_bev is [N]
        for i in range(boxes_a.shape[0]):
            # Mock BEV overlap computation - just use first two dimensions
            dx = abs(boxes_a[i, 0] - boxes_b[i, 0])
            dy = abs(boxes_a[i, 1] - boxes_b[i, 1])
            # Simple mock overlap: decrease with distance
            overlaps_bev[i] = max(0.0, 1.0 - (dx + dy) / 10.0)

# Mock the external CUDA module
iou3d_nms_diff_cuda = MockIou3dNmsDiffCuda()

class MockNumericalJacobian:
    @staticmethod
    def get_numerical_jacobian(fn, inputs, target_input, eps=0.001):
        # Returns a Jacobian matrix of shape [output_dim, input_dim]
        with torch.no_grad():
            output = fn(inputs)
            output_dim = output.shape[0]
            input_dim = target_input.numel()
            jacobian = torch.zeros(output_dim, input_dim)
            
            # Simple numerical differentiation
            for i in range(input_dim):
                target_input_plus = target_input.clone()
                target_input_plus.view(-1)[i] += eps
                inputs_plus = (target_input_plus, inputs[1])
                output_plus = fn(inputs_plus)
                
                target_input_minus = target_input.clone()
                target_input_minus.view(-1)[i] -= eps
                inputs_minus = (target_input_minus, inputs[1])
                output_minus = fn(inputs_minus)
                
                jacobian[:, i] = (output_plus - output_minus) / (2 * eps)
        
        return jacobian

numerical_jaccobian = MockNumericalJacobian()

# Create valid input tensors for the function
# boxes format: [x, y, z, length, width, height, rotation] (7 dimensions)
boxes_a = torch.randn(2, 7, dtype=torch.float32)
boxes_b = torch.randn(2, 7, dtype=torch.float32)

# Ensure dimensions make sense (lengths, widths, heights should be positive)
boxes_a[:, 3:6] = torch.abs(boxes_a[:, 3:6]) + 0.1
boxes_b[:, 3:6] = torch.abs(boxes_b[:, 3:6]) + 0.1

ARGS = {
    'boxes_a': boxes_a,
    'boxes_b': boxes_b,
}
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 BoxesIou3dDifferentiableFunction — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a BoxesIou3dDifferentiableFunction 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
E2ESigmoidFocalClassificationLoss Ran
ran on a synthesized input · constructed an object; no result was computed
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("3edc8ae9bf725ed7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, E2ESigmoidFocalClassificationLoss: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for input and target
batch_size = 2
num_anchors = 4
num_classes = 3

# Input: logits for each class (B, #anchors, #classes)
input_tensor = torch.randn(batch_size, num_anchors, num_classes)

# Target: one-hot encoded classification targets (B, #anchors, #classes)
target_tensor = torch.zeros(batch_size, num_anchors, num_classes)
# Set some random positions to 1 for one-hot encoding
target_tensor[0, 0, 0] = 1.0
target_tensor[0, 1, 1] = 1.0
target_tensor[1, 2, 2] = 1.0
target_tensor[1, 3, 0] = 1.0

ARGS = {
    'gamma': 2.0,
    'alpha': 0.25,
    'reduction': 'mean'
}
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 E2ESigmoidFocalClassificationLoss — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a E2ESigmoidFocalClassificationLoss 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
GroundTruthProcessor Ran
ran on a synthesized input · constructed an object; no result was computed
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("5f96efdd4664c13e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, GroundTruthProcessor: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

class Task:
    def __init__(self, class_names):
        self.class_names = class_names

class GTProcessorConfig:
    def __init__(self):
        self.tasks = [
            Task(['car', 'pedestrian']),
            Task(['bicycle', 'motorcycle'])
        ]
        self.mapping = {
            'car': 0,
            'pedestrian': 1,
            'bicycle': 2,
            'motorcycle': 3
        }

gt_processor_cfg = GTProcessorConfig()

ARGS = {
    'gt_processor_cfg': gt_processor_cfg
}
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 GroundTruthProcessor — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a GroundTruthProcessor 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
OneNetSeqHeadTSC Ran
ran on a synthesized input · constructed an object; no result was computed
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("26bb8484040b20a5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, OneNetSeqHeadTSC: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Mock common_utils module since it's used in the function signatures
class common_utils:
    @staticmethod
    def rotate_points_along_z(points, angle):
        """
        Args:
            points: (B, N, 3 + C)
            angle: (B), angle along z-axis, angle increases x ==> y
        Returns:
        """
        cosa = torch.cos(angle)
        sina = torch.sin(angle)
        zeros = angle.new_zeros(points.shape[0])
        ones = angle.new_ones(points.shape[0])
        rot_matrix = torch.stack((cosa, sina, zeros, -sina, cosa, zeros, zeros, zeros, ones), dim=1).view(-1, 3, 3).float()
        points_rot = torch.matmul(points[:, :, 0:3], rot_matrix)
        points_rot = torch.cat((points_rot, points[:, :, 3:]), dim=-1)
        return points_rot

ARGS = {
    'in_channels': 64,
    'heads': {
        'kernel_size': 3,
        'num_classes': 5,
        'head_channels': 32,
        'init_bias': -2.19,
        'template_box': [4.0, 2.0, 1.5],
        'pc_range': [0.0, -40.0, -3.0, 70.0, 40.0, 1.0],
        'voxel_size': [0.1, 0.1, 0.2],
        'offset_grid': torch.zeros(2, 1, 1, 2),
    }
}
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 OneNetSeqHeadTSC — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a OneNetSeqHeadTSC 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
SmoothL1Loss Ran
ran on a synthesized input
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("6ca556054d761a3c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

# INIT_ARGS for SmoothL1Loss.__init__(sigma, reduction='mean')
INIT_ARGS = {
    'sigma': 1.5,
    'reduction': 'mean'
}

# ARGS for SmoothL1Loss.forward(x)
ARGS = {
    'x': torch.randn(3, 5)  # batch_size=3, feature_dim=5
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 1, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
box_to_surface_bev Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("887b3486b120474c")
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 3, dtype family float, shape (2, 4, 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 torch
import numpy as np
import types as _types

torch.manual_seed(0)

def check_numpy_to_torch(x):
    if isinstance(x, np.ndarray):
        return (torch.from_numpy(x).float(), True)
    return (x, False)

def rotate_points_along_z(points, angle):
    """
    Args:
        points: (B, N, 3 + C)
        angle: (B), angle along z-axis, angle increases x ==> y
    Returns:

    """
    points, is_numpy = check_numpy_to_torch(points)
    angle, _ = check_numpy_to_torch(angle)
    cosa = torch.cos(angle)
    sina = torch.sin(angle)
    zeros = angle.new_zeros(points.shape[0])
    ones = angle.new_ones(points.shape[0])
    rot_matrix = torch.stack((cosa, sina, zeros, -sina, cosa, zeros, zeros, zeros, ones), dim=1).view(-1, 3, 3).float()
    points_rot = torch.matmul(points[:, :, 0:3], rot_matrix)
    points_rot = torch.cat((points_rot, points[:, :, 3:]), dim=-1)
    return points_rot.numpy() if is_numpy else points_rot

common_utils = _types.SimpleNamespace(**{n: globals()[n] for n in ['check_numpy_to_torch', 'rotate_points_along_z'] if n in globals()})

def box_to_surface_bev(box, angle, raw_xy, box_scale_ratio=1.0):
    """
       x ---2--- x
      1         3
    x ---4--- x
    Args:
        box:  (bs, 3) [dx, dy, dz]
        angle: B
        raw_xy: B, 2

    Returns:
    """
    template = box.new_tensor(([1, 0, 0], [0, 1, 0], [-1, 0, 0], [0, -1, 0])) / 2 * box_scale_ratio
    surface_pts = box.unsqueeze(dim=1) * template.unsqueeze(dim=0)
    surface_pts = common_utils.rotate_points_along_z(surface_pts, angle).view(-1, 4, 3)[..., :2].contiguous()
    surface_pts = surface_pts + raw_xy.unsqueeze(dim=1)
    return surface_pts

# Setup arguments
batch_size = 2
box = torch.randn(batch_size, 3)
angle = torch.randn(batch_size)
raw_xy = torch.randn(batch_size, 2)

ARGS = {
    "box": box,
    "angle": angle,
    "raw_xy": raw_xy,
    "box_scale_ratio": 1.0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 4, 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
check_numpy_to_torch Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from livox-sdk/livox_detection
pointer only · get_code("5c4bac81583c34e1")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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_world_size Ran
ran on a synthesized input
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("e78ed49296e4bb70")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function get_world_size() takes no parameters
# It only calls is_distributed() which checks dist.is_available() and dist.is_initialized()
# Since we're in a non-distributed environment, it will return 1

ARGS = {}
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 int, ndim 0, dtype family int, 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
is_distributed Ran
ran on a synthesized input
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("30d203cc642954ed")
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; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.distributed as dist

torch.manual_seed(0)

# The function is_distributed() takes no parameters
# It checks:
# 1. dist.is_available() - returns False if torch.distributed is not available
# 2. dist.is_initialized() - returns False if distributed training is not initialized
# Since this is a CPU-only environment and we're not setting up actual distributed training,
# the function will return False (either because dist is not available or not initialized)

ARGS = {}
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 bool, ndim 0, dtype family float — no output shape is stated: observe_signature.py could not observe it (no_inputs)
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
label_to_one_hot Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("0572d4f99dca2d49")
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 3, dtype family float, shape (2, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Analyze the function:
# - label: a tensor that gets unsqueezed(-1), then used with scatter_
# - pred: a tensor with .size() that returns (bs, querys, cls)
#   and has a .device attribute

# The function expects:
# - label to be indexable for scatter_, shape should be (bs, querys)
# - pred to be shape (bs, querys, cls)

# Create tensors with distinct dimensions for clarity
bs = 2      # batch size
querys = 3  # query dimension
cls = 4     # class dimension

# label should be (bs, querys) with integer values in range [0, cls]
label = torch.randint(0, cls + 1, (bs, querys), dtype=torch.long)

# pred should be (bs, querys, cls)
pred = torch.randn(bs, querys, cls, dtype=torch.float32)

ARGS = {
    "label": label,
    "pred": pred,
}
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
limit_period_torch Ran
ran, honoured the contract we drafted
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("017ecc57a0af8b57")
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, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import torch

torch.manual_seed(0)

# Create a small tensor for val parameter
val = torch.randn(2, 4)

# offset parameter - default is 0, but we can provide a value
offset = 0.5

# period parameter - default is 2*pi, but we can provide a different value
period = 2 * np.pi

ARGS = {
    "val": val,
    "offset": offset,
    "period": period
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rotate_points_along_z Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
this paper's copy was not recorded; identical code first harvested from livox-sdk/livox_detection
pointer only · get_code("df17a2b345c6cccd")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

template_to_surface_bev Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("393718acf2b116e2")
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 3, dtype family float, shape (2, 4, 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 torch
import numpy as np
import types as _types

torch.manual_seed(0)

def check_numpy_to_torch(x):
    if isinstance(x, np.ndarray):
        return (torch.from_numpy(x).float(), True)
    return (x, False)

def rotate_points_along_z(points, angle):
    """
    Args:
        points: (B, N, 3 + C)
        angle: (B), angle along z-axis, angle increases x ==> y
    Returns:

    """
    points, is_numpy = check_numpy_to_torch(points)
    angle, _ = check_numpy_to_torch(angle)
    cosa = torch.cos(angle)
    sina = torch.sin(angle)
    zeros = angle.new_zeros(points.shape[0])
    ones = angle.new_ones(points.shape[0])
    rot_matrix = torch.stack((cosa, sina, zeros, -sina, cosa, zeros, zeros, zeros, ones), dim=1).view(-1, 3, 3).float()
    points_rot = torch.matmul(points[:, :, 0:3], rot_matrix)
    points_rot = torch.cat((points_rot, points[:, :, 3:]), dim=-1)
    return points_rot.numpy() if is_numpy else points_rot

def template_to_surface_bev(template_box, angle, raw_xy, box_scale_ratio=1.0):
    """
       x ---2--- x
      1         3
    x ---4--- x
    Args:
        template_box:  (3) [dx, dy, dz]
        angle: B
        raw_xy: B, 2

    Returns:
    """
    bs = angle.shape[0]
    template = template_box.new_tensor(([1, 0, 0], [0, 1, 0], [-1, 0, 0], [0, -1, 0])) / 2 * box_scale_ratio
    surface_pts = template_box.unsqueeze(dim=0) * template
    surface_pts = surface_pts.unsqueeze(dim=0)
    surface_pts = surface_pts.repeat(bs, 1, 1)
    surface_pts = common_utils.rotate_points_along_z(surface_pts, angle).view(-1, 4, 3)[..., :2].contiguous()
    surface_pts = surface_pts + raw_xy.unsqueeze(dim=1)
    return surface_pts

common_utils = _types.SimpleNamespace(**{n: globals()[n] for n in ['check_numpy_to_torch', 'rotate_points_along_z'] if n in globals()})

ARGS = {
    'template_box': torch.tensor([2.0, 3.0, 1.5]),
    'angle': torch.tensor([0.1, 0.2]),
    'raw_xy': torch.tensor([[1.0, 2.0], [3.0, 4.0]]),
    'box_scale_ratio': 1.0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 4, 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
E2ESeqHead Not yet run
unverified (harvested, no recorded run)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("97fa820cfae5e65d")
How it ran

no invocation recorded.

IOU3DLoss Not yet run
unverified (harvested, no recorded run)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("9e12fe042064e780")
How it ran

no invocation recorded.

OneNetSeqHead Not yet run
unverified (harvested, no recorded run)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("ffffba4bad43d9ed")
How it ran

no invocation recorded.

SetCriterion Not yet run
unverified (harvested, no recorded run)
ocnflag/point2seq/pcdet/models/dense_heads/e2e_seq_head.py
code served (permissive licence) · get_code("91ff85395933dcaf")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 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
check_numpy_to_torch
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
2303.15286 (2023-03)
Unsupervised Adaptation from Repeated Traversals for Autonomous Driving
2412.01812 (2024-12)
V2XPnP: Vehicle-to-Everything Spatio-Temporal Fusion for Multi-Agent Perception and Prediction
rotate_points_along_z
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking

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 a simple and effective framework, named Point2Seq, for 3D object detection from point clouds. In contrast to previous methods that normally predict attributes of 3D objects all at once, we expressively model the interdependencies between attributes of 3D objects, which in turn enables a better detection accuracy. Specifically, we view each 3D object as a sequence of words and reformulate the 3D object detection task as decoding words from 3D scenes in an auto-regressive manner. We further propose a lightweight scene-to-sequence decoder that can auto-regressively generate words conditioned on features from a 3D scene as well as cues from the preceding words. The predicted words eventually constitute a set of sequences that completely describe the 3D objects in the scene, and all the predicted sequences are then automatically assigned to the respective ground truths through similarity-based sequence matching. Our approach is conceptually intuitive and can be readily plugged upon most existing 3D-detection backbones without adding too much computational overhead; the sequential decoding paradigm we proposed, on the other hand, can better exploit information from complex 3D scenes with the aid of preceding predicted words. Without bells and whistles, our method significantly outperforms previous anchor-and center-based 3D object detection frameworks, yielding the new state of the art on the challenging ONCE dataset as well as the Waymo Open Dataset. Code is available at https://github.com/ ocNflag/point2seq.

For agents

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

get_harvested_code_for_paper("2203.13394")
get_code_for_paper("2203.13394")
have("2203.13394")

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

curl https://syntology.ai/api/ran/2203.13394.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.13394.svg)](https://syntology.ai/paper/2203.13394)

Connect an agent — have() is free.