SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.15670 · NeurIPS · 2023

Leveraging Vision-Centric Multi-Modal Expertise for 3D Object Detection

Hongyang Li, Jingdong Wang, Yu Qiao, Wenhai Wang, Zhiqi Li, Chonghao Sima, Linyan Huang

arXiv · PDF · Open in the Atlas

Code that ran

4 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 3 failures of our instrument; 0 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/2310.15670.json · badge: /api/ran/2310.15670.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
OpenDriveLab/Birds-eye-view-Perception canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
FunctionStatusWhere it lives
gen_dx_bx Ran
ran, violated the contract we drafted
OpenDriveLab/Birds-eye-view-Perception/nuScenes_playground/VCD/mmdet3d/models/necks/view_transformer.py
code served (permissive licence) · get_code("f6a173f4830f978c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Each bound is a list with [min, max, step]
# The function expects three bounds: xbound, ybound, zbound
xbound = [0.0, 10.0, 0.5]
ybound = [-5.0, 5.0, 0.25]
zbound = [0.0, 20.0, 1.0]

ARGS = {
    "xbound": xbound,
    "ybound": ybound,
    "zbound": zbound,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
horizontal_flip_image_multiview Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
OpenDriveLab/Birds-eye-view-Perception/bev_toolbox/data_aug/functional.py
code served (permissive licence) · get_code("d6f1a151dcb09c1f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 4, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a list of small multi-view images
# Each image is [H, W, 3] where H=4, W=6, C=3
img1 = np.random.randint(0, 256, size=(4, 6, 3), dtype=np.uint8)
img2 = np.random.randint(0, 256, size=(4, 6, 3), dtype=np.uint8)
img3 = np.random.randint(0, 256, size=(4, 6, 3), dtype=np.uint8)

imgs = [img1, img2, img3]

ARGS = {
    "imgs": imgs
}
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 list, ndim 4, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
scale_image_multiple_view Ran
ran; our contract draft was wrong, not the code
OpenDriveLab/Birds-eye-view-Perception/bev_toolbox/data_aug/functional.py
code served (permissive licence) · get_code("6b894132bebbea7e")
How it ran

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

import numpy as np
import cv2
from typing import List, Tuple

# Create sample images for multiple cameras
# Using small 32x48 images (H x W x C format)
img1 = np.random.randint(0, 255, (32, 48, 3), dtype=np.uint8)
img2 = np.random.randint(0, 255, (32, 48, 3), dtype=np.uint8)
imgs = [img1, img2]

# Create intrinsic matrices (4x4 each)
# Intrinsic matrix transforms from camera to image coordinates
cam_intrinsic1 = np.array([
    [500, 0, 24, 0],
    [0, 500, 16, 0],
    [0, 0, 1, 0],
    [0, 0, 0, 1]
], dtype=np.float32)

cam_intrinsic2 = np.array([
    [480, 0, 24, 0],
    [0, 480, 16, 0],
    [0, 0, 1, 0],
    [0, 0, 0, 1]
], dtype=np.float32)

cam_intrinsics = [cam_intrinsic1, cam_intrinsic2]

# Create extrinsic matrices (4x4 each)
# Extrinsic matrix transforms from lidar to camera
cam_extrinsic1 = np.array([
    [1, 0, 0, 0.1],
    [0, 1, 0, 0.2],
    [0, 0, 1, 0.5],
    [0, 0, 0, 1]
], dtype=np.float32)

cam_extrinsic2 = np.array([
    [1, 0, 0, -0.1],
    [0, 1, 0, 0.2],
    [0, 0, 1, 0.5],
    [0, 0, 0, 1]
], dtype=np.float32)

cam_extrinsics = [cam_extrinsic1, cam_extrinsic2]

# Create lidar2img transformation matrices (4x4 each)
# This combines extrinsic and intrinsic transformations
lidar2img1 = cam_intrinsic1 @ cam_extrinsic1
lidar2img2 = cam_intrinsic2 @ cam_extrinsic2
lidar2img = [lidar2img1, lidar2img2]

# Random scale factor
rand_scale = 0.5

# Interpolation mode
interpolation = 'bilinear'

ARGS = {
    'imgs': imgs,
    'cam_intrinsics': cam_intrinsics,
    'cam_extrinsics': cam_extrinsics,
    'lidar2img': lidar2img,
    'rand_scale': rand_scale,
    'interpolation': interpolation
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
vertical_flip_image_multiview Ran
ran; our contract draft was wrong, not the code
OpenDriveLab/Birds-eye-view-Perception/bev_toolbox/data_aug/functional.py
code served (permissive licence) · get_code("2bbcfd0ababb1e92")
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. 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 multi-view images with distinct dimensions for clarity
# Each image should have shape [H, W, 3] as specified in the docstring

# Camera 1: 4 height x 5 width x 3 channels
img1 = np.random.randint(0, 256, size=(4, 5, 3), dtype=np.uint8)

# Camera 2: 6 height x 7 width x 3 channels (distinct size)
img2 = np.random.randint(0, 256, size=(6, 7, 3), dtype=np.uint8)

# Camera 3: 8 height x 9 width x 3 channels (distinct size)
img3 = np.random.randint(0, 256, size=(8, 9, 3), dtype=np.uint8)

imgs = [img1, img2, img3]

ARGS = {
    "imgs": imgs
}
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 list — a list 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

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 1 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
gen_dx_bx 1 other paper
1 later
2602.04240 (2026-02)
SPOT-Occ: Sparse Prototype-guided Transformer for Camera-based 3D Occupancy Prediction
horizontal_flip_image_multiview 1 other paper
1 dated earlier
2304.03105 (2023-04)
Geometric-aware Pretraining for Vision-centric 3D Object Detection
scale_image_multiple_view 1 other paper
1 dated earlier
2304.03105 (2023-04)
Geometric-aware Pretraining for Vision-centric 3D Object Detection
vertical_flip_image_multiview 1 other paper
1 dated earlier
2304.03105 (2023-04)
Geometric-aware Pretraining for Vision-centric 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

Current research is primarily dedicated to advancing the accuracy of camera-only 3D object detectors (apprentice) through the knowledge transferred from LiDARor multi-modal-based counterparts (expert). However, the presence of the domain gap between LiDAR and camera features, coupled with the inherent incompatibility in temporal fusion, significantly hinders the effectiveness of distillation-based enhancements for apprentices. Motivated by the success of uni-modal distillation, an apprentice-friendly expert model would predominantly rely on camera features, while still achieving comparable performance to multi-modal models. To this end, we introduce VCD, a framework to improve the camera-only apprentice model, including an apprentice-friendly multi-modal expert and temporal-fusion-friendly distillation supervision. The multi-modal expert VCD-E adopts an identical structure as that of the camera-only apprentice in order to alleviate the feature disparity, and leverages LiDAR input as a depth prior to reconstruct the 3D scene, achieving the performance on par with other heterogeneous multi-modal experts. Additionally, a fine-grained trajectory-based distillation module is introduced with the purpose of individually rectifying the motion misalignment for each object in the scene. With those improvements, our camera-only apprentice VCD-A sets new state-of-the-art on nuScenes with a score of 63.1% NDS. The code will be released at https://github.com/OpenDriveLab/Birds-eye-view-Perception.

For agents

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

get_harvested_code_for_paper("2310.15670")
get_code_for_paper("2310.15670")
have("2310.15670")

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

curl https://syntology.ai/api/ran/2310.15670.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.15670.svg)](https://syntology.ai/paper/2310.15670)

Connect an agent — have() is free.