SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · aaai_28577 · AAAI · 2024

SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection

Dacheng Tao, Jing Zhang, Qiming Zhang, Zhe Chen, Shanshan Zhao, Haimei Zhao

Open in the Atlas

Code that ran

10 of 10 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; 2 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 4 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/aaai_28577.json · badge: /api/ran/aaai_28577.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
ViTAE-Transformer/SimDistill canonical (classifier label)
classifier basis: census
10 ran · 0 no recorded run
FunctionStatusWhere it lives
aligned_3d_nms Ran
ran on a synthesized input
ViTAE-Transformer/SimDistill/mmdet3d/core/post_processing/box3d_nms.py
code served (permissive licence) · get_code("0079ed3e99e4048d")
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 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 torch

torch.manual_seed(0)

# Create small 3D bounding boxes with shape [n, 6]
# Each box is [x1, y1, z1, x2, y2, z2]
boxes = torch.tensor([
    [0.0, 0.0, 0.0, 2.0, 2.0, 2.0],
    [1.0, 1.0, 1.0, 3.0, 3.0, 3.0],
], dtype=torch.float32)

# Scores for each box
scores = torch.tensor([0.9, 0.7], dtype=torch.float32)

# Class labels for each box
classes = torch.tensor([0, 0], dtype=torch.long)

# IoU threshold for NMS
thresh = 0.5

ARGS = {
    "boxes": boxes,
    "scores": scores,
    "classes": classes,
    "thresh": thresh,
}
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 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
box_camera_to_lidar Ran
ran, honoured the contract we drafted
ViTAE-Transformer/SimDistill/mmdet3d/core/bbox/box_np_ops.py
code served (permissive licence) · get_code("96c8084488bd023b")
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, 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 numpy as np

np.random.seed(0)

# Create small sample data for box_camera_to_lidar function
# data shape should be [N, 7] where N is number of boxes
# Each box has: [x, y, z, l, h, w, r] (3 position + 3 dimensions + 1 rotation)
data = np.array([
    [5.0, 10.0, 2.0, 3.5, 1.6, 1.8, 0.5],
    [8.0, 12.0, 1.5, 3.2, 1.5, 1.7, 1.2]
], dtype=np.float32)

# r_rect: [4, 4] matrix to project from CAM2 to CAM0
# This is typically the camera rectification matrix
r_rect = np.array([
    [0.9999, -0.0050, 0.0095, 0.0],
    [0.0050, 0.9999, -0.0010, 0.0],
    [-0.0095, 0.0010, 0.9999, 0.0],
    [0.0, 0.0, 0.0, 1.0]
], dtype=np.float32)

# velo2cam: [4, 4] matrix to project from lidar/velodyne to camera coordinate
# This is the extrinsic calibration matrix
velo2cam = np.array([
    [0.0125, -0.9992, 0.0359, -0.0430],
    [-0.2357, -0.0381, -0.9714, -0.2330],
    [0.9719, 0.0106, -0.2360, -0.0144],
    [0.0, 0.0, 0.0, 1.0]
], dtype=np.float32)

ARGS = {
    'data': data,
    'r_rect': r_rect,
    'velo2cam': velo2cam
}
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 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
camera_to_lidar Ran
ran, honoured the contract we drafted
ViTAE-Transformer/SimDistill/mmdet3d/core/bbox/box_np_ops.py
code served (permissive licence) · get_code("9e22984820c39868")
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

# Create small test data for camera_to_lidar function
# points: shape [N, 3] - points in camera coordinate
np.random.seed(0)
points = np.random.randn(2, 3).astype(np.float32)

# r_rect: shape [4, 4] - matrix to project points from specific camera coord to CAM0
# Create a valid 4x4 transformation matrix
r_rect = np.eye(4, dtype=np.float32)
r_rect[0, 1] = 0.1
r_rect[1, 2] = 0.05

# velo2cam: shape [4, 4] - matrix to project points from camera coord to lidar coord
# Create a valid 4x4 transformation matrix
velo2cam = np.eye(4, dtype=np.float32)
velo2cam[0, 0] = 0.9
velo2cam[1, 1] = 0.95
velo2cam[2, 2] = 1.0

ARGS = {
    "points": points,
    "r_rect": r_rect,
    "velo2cam": velo2cam,
}
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 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
circle_nms Ran
ran on a synthesized input
ViTAE-Transformer/SimDistill/mmdet3d/core/post_processing/box3d_nms.py
code served (permissive licence) · get_code("a8007c7a213459b2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create detection results with shape [N, 3]
# N=4 detections, each with [x, y, confidence_score]
dets = np.array([
    [10.0, 20.0, 0.9],
    [11.0, 21.0, 0.7],
    [50.0, 60.0, 0.8],
    [100.0, 150.0, 0.6]
], dtype=np.float64)

# Threshold for circular NMS (squared distance threshold in bird's eye view)
thresh = 4.0

# Maximum number of detections to keep after NMS
post_max_size = 83

ARGS = {
    "dets": dets,
    "thresh": thresh,
    "post_max_size": post_max_size
}
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 list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
corners_nd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ViTAE-Transformer/SimDistill/mmdet3d/core/bbox/box_np_ops.py
code served (permissive licence) · get_code("5b1dacd1f125c08f")
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 (2, 8, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create valid arguments for corners_nd function
dims = np.array([[2.0, 3.0, 1.5], [4.0, 2.5, 3.0]], dtype=np.float32)

origin = 0.5

ARGS = {
    "dims": dims,
    "origin": origin,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 8, 3) — shape (2, 8, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
denormalize_bbox Ran
ran; our contract draft was wrong, not the code
ViTAE-Transformer/SimDistill/mmdet3d/core/bbox/util.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from sranc3/m-bev, is pointer only (licence: NONE) · get_code("38e90d12ef38fe79")
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 sranc3/m-bev, 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.

draw_heatmap_gaussian Ran
ran on a synthesized input
ViTAE-Transformer/SimDistill/mmdet3d/core/utils/gaussian.py
code served (permissive licence) · get_code("a7cde40ccc6422e2")
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 (10, 12) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create a small heatmap tensor (height=10, width=12)
heatmap = torch.zeros((10, 12), dtype=torch.float32)

# Create center coordinates as a tensor
center = torch.tensor([5.0, 4.0])

# Set radius for the gaussian
radius = 3

# Set k (multiplier)
k = 1

ARGS = {
    "heatmap": heatmap,
    "center": center,
    "radius": radius,
    "k": k
}
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 (10, 12) — shape (10, 12) 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
gaussian_2d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ViTAE-Transformer/SimDistill/mmdet3d/core/utils/gaussian.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from yifanlu0227/HEAL, is pointer only (licence: NOASSERTION) · get_code("ea036153b550642a")
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 yifanlu0227/HEAL, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

gaussian_radius Ran
ran on a synthesized input
ViTAE-Transformer/SimDistill/mmdet3d/core/utils/gaussian.py
code served (permissive licence) · get_code("73e8316be86a800d")
How it ran

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

import torch

torch.manual_seed(0)

# det_size is a tuple of two torch.Tensor values representing height and width
height = torch.tensor(10.0)
width = torch.tensor(8.0)
det_size = (height, width)

# min_overlap is a float parameter
min_overlap = 0.5

ARGS = {
    "det_size": det_size,
    "min_overlap": min_overlap,
}
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 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
normalize_bbox Ran
ran; our contract draft was wrong, not the code
ViTAE-Transformer/SimDistill/mmdet3d/core/bbox/util.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from sranc3/m-bev, is pointer only (licence: NONE) · get_code("e6ab81c2ff932920")
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 sranc3/m-bev, 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.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 10 are also present in other papers whose code we hold (10 with at least one paper dated earlier, 0 the same month, 4 later) and 0 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
box_camera_to_lidar 9 other papers
7 dated earlier, 1 later, 1 same year, order not determinable
2104.10956 (2021-04)
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection
136990210 (2022 (year only))
FH-Net: A Fast Hierarchical Network for Scene Flow Estimation on Real-world Point Clouds
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
and 6 more not named here
camera_to_lidar 9 other papers
7 dated earlier, 1 later, 1 same year, order not determinable
2104.10956 (2021-04)
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection
136990210 (2022 (year only))
FH-Net: A Fast Hierarchical Network for Scene Flow Estimation on Real-world Point Clouds
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
and 6 more not named here
corners_nd 8 other papers
6 dated earlier, 1 later, 1 same year, order not determinable
2104.10956 (2021-04)
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
and 5 more not named here
normalize_bbox 7 other papers
7 dated earlier
2110.06923 (2021-10)
Object DGCNN: 3D Object Detection using Dynamic Graphs
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
and 4 more not named here
denormalize_bbox 6 other papers
6 dated earlier
2110.06923 (2021-10)
Object DGCNN: 3D Object Detection using Dynamic Graphs
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
and 3 more not named here
gaussian_2d 6 other papers
4 dated earlier, 1 later, 1 same year, order not determinable
2006.11275 (2020-06)
Center-based 3D Object Detection and Tracking
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
and 3 more not named here
aligned_3d_nms 4 other papers
4 dated earlier
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
2210.05593 (2022-10)
Prototypical VoteNet for Few-Shot 3D Point Cloud Object Detection
and 1 more not named here
circle_nms 4 other papers
4 dated earlier
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
2210.05593 (2022-10)
Prototypical VoteNet for Few-Shot 3D Point Cloud Object Detection
and 1 more not named here
draw_heatmap_gaussian 3 other papers
3 dated earlier
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
2303.16818 (2023-03)
SimDistill: Simulated Multi-modal Distillation for BEV 3D Object Detection
gaussian_radius 3 other papers
3 dated earlier
2203.10642 (2022-03)
FUTR3D: A Unified Sensor Fusion Framework for 3D Detection
2205.13542 (2022-05)
BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
2303.16818 (2023-03)
SimDistill: Simulated Multi-modal Distillation for BEV 3D Object Detection

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

Repositories linked to this paper

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

Abstract

Multi-view camera-based 3D object detection has become popular due to its low cost, but accurately inferring 3D geometry solely from camera data remains challenging and may lead to inferior performance. Although distilling precise 3D geometry knowledge from LiDAR data could help tackle this challenge, the benefits of LiDAR information could be greatly hindered by the significant modality gap between different sensory modalities. To address this issue, we propose a Simulated multi-modal Distillation (SimDistill) method by carefully crafting the model architecture and distillation strategy. Specifically, we devise multi-modal architectures for both teacher and student models, including a LiDARcamera fusion-based teacher and a simulated fusion-based student. Owing to the "identical" architecture design, the student can mimic the teacher to generate multi-modal features with merely multi-view images as input, where a geometry compensation module is introduced to bridge the modality gap. Furthermore, we propose a comprehensive multimodal distillation scheme that supports intra-modal, crossmodal, and multi-modal fusion distillation simultaneously in the Bird's-eye-view space. Incorporating them together, our SimDistill can learn better feature representations for 3D object detection while maintaining a cost-effective camera-only deployment. Extensive experiments validate the effectiveness and superiority of SimDistill over state-of-the-art methods, achieving an improvement of 4.8% mAP and 4.1% NDS over the baseline detector. The source code will be released at https://github.com/ViTAE-Transformer/SimDistill.

For agents

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

get_harvested_code_for_paper("aaai_28577")
get_code_for_paper("aaai_28577")
have("aaai_28577")

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

curl https://syntology.ai/api/ran/aaai_28577.json

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

[![Syntology run record](https://syntology.ai/api/ran/aaai_28577.svg)](https://syntology.ai/paper/aaai_28577)

Connect an agent — have() is free.