SafeAug: Safety-Critical Driving Data Augmentation from Naturalistic Datasets

Xuan Di; Yunlong Li; Zhaobin Mo

REVIEW 4 major objections 6 minor 26 references

SafeAug: Safety-Critical Driving Data Augmentation from Naturalistic Datasets

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Editing real KITTI images to move the front vehicle half a body length closer, and scaling the acceleration label by 1.5, reduces a CNN's error on safety-critical acceleration prediction below SMOGN and importance sampling.

desk verdict The central claim is confounded: the image edit and the 1.5x label rescaling are never separated, so the reported safety-critical RMSE gain cannot be attributed to the geometric augmentation. read the letter →

arxiv 2501.02143 v1 pith:ACLLGEF4 submitted 2025-01-03 cs.CV cs.LG

Zhaobin Mo , Yunlong Li , Xuan Di This is my paper · ORCID

classification cs.CVcs.LG

keywords dataaugmentationsafety-criticaldrivingnaturalisticdatasetsdepthestimation3DpointcloudeditingvehicleaccelerationpredictionKITTIautonomous

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Safety-critical driving situations are rare in naturalistic datasets, so autonomous-driving models are usually trained with simulated or synthetically generated hazard images. This paper proposes instead to manufacture hazard from real footage: take a KITTI image, detect the vehicle ahead, reconstruct the scene as a 3D point cloud, move that vehicle half a body length closer, re-project the edited scene into a new image, and multiply the recorded acceleration label by 1.5. The claim is that a basic convolutional neural network trained on roughly 2,000 original images plus about 200 such edited images predicts vehicle acceleration on the most safety-critical KITTI frames with lower error than training on the original set or on the original set augmented by SMOGN or importance sampling. The method's stated purpose is to keep the authenticity of naturalistic imagery while shifting the data distribution toward hazard.

What carries the argument

The enabling mechanism is an image-to-3D-to-image editing pipeline. YOLOv5, an object detector, finds the vehicle directly in front; Depth-Anything, a monocular depth estimator, supplies a dense depth map; Open3D fuses the image and depth map into a point cloud; the front vehicle's points are shifted toward the camera by half a body length; and the modified cloud is projected back into a 2D image. In parallel, the acceleration label is rescaled as $A_{aug} = 1.5 \times A_{orig}$, so each new sample pairs a visually closer vehicle with a proportionally stronger deceleration label.

What would settle it

Train the same CNN on the original KITTI frames with only the acceleration labels multiplied by 1.5 and no geometric edit. If safety-critical RMSE on the held-out frames falls to about 1.69, the reported gain does not require the image augmentation; a second check is to compute the deceleration needed to stop within the reduced gap at the recorded speed and compare it with 1.5 times the original label.

Watch

Extended reading notes

Core claim

The central discovery claimed is that a geometric image edit plus a proportional label change can make a naturalistic driving dataset more useful for safety-critical prediction without sacrificing general performance. In the reported experiments, the augmented dataset achieves safety-critical RMSE/MAE of 1.6923/1.4312, versus 1.8725/1.6504 for the original data, 1.6966/1.4883 for SMOGN, and 1.7934/1.5635 for importance sampling; on the complete dataset it achieves 0.2039/0.1132, versus 0.2416/0.1217 for the original data. The authors take this as evidence that their authenticity-preserving augmentation improves the model's handling of hazardous situations while also slightly improving ordinary-condition predictions.

Load-bearing premise

The load-bearing premise is that the 1.5 times acceleration rescaling is the physically correct ground-truth label for the closer-vehicle scene; if that scaling is arbitrary, the reported safety-critical improvement could come entirely from the relabeling, with no contribution from the geometric image edit.

Editorial extensions

If this is right

A model trained on the mixed original-plus-augmented set predicts acceleration on the 10 percent most safety-critical KITTI frames with lower RMSE and MAE than the original set, SMOGN, or importance sampling.
The same augmented set also improves complete-dataset RMSE/MAE, so the safety-critical gain does not come at the cost of ordinary driving accuracy.
Because the edited images are re-projections of real camera frames rather than synthetic renders, the method retains more of the naturalistic image distribution than simulated or generated alternatives.
The hazard level is controlled by the shift distance (half a body length) and the label multiplier (1.5), so the pipeline can generate a range of criticality from the same source images.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Algorithm 4 rescales the label to 1.5 times the original without a braking-dynamics derivation, and the reported experiment never ablates the image edit away from the relabeling; retraining with only the relabeled acceleration on the original images would show how much of the safety-critical gain is attributable to the geometric change.
The half-body-length shift re-projects the point cloud without regenerating occluded background, so the visual edit stays natural only for frames where the initial gap is large enough and the line of sight is clear; frames with very close or heavily occluded lead cars are likely to show artifacts.
A physically grounded alternative would set the new label from the required deceleration at the shortened gap rather than a fixed 1.5 multiplier, and the augmented labels could then be checked against real emergency-braking telemetry.
If the relabeling alone reproduces most of the reported gain, the practical recipe would shrink to re-weighting or relabeling existing near-miss frames, with no need for 3D reconstruction.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The core claim doesn't survive contact with the experiment: the reported gain is a tie between moving the car and relabeling the acceleration, and the paper never unties it. Algorithm 4 changes two variables at once—the image is modified and Aaug is set to 1.5×Aorig—while the test set is the 10% of frames with the most negative accelerations. Shifting training labels further into that tail can reduce RMSE on that tail regardless of whether the image edit conveys any useful visual information. That is a textbook confound, and it is load-bearing: the paper explicitly attributes the improvement to the geometric augmentation.

What is genuinely nice here is the direction: using naturalistic images and depth-based geometric warping to create near-following hazards without a generator. The pipeline is clearly described, and the front-vehicle selection via the center-line and bounding-box area is sensible. If the confound were resolved, this could be a practical tool for AV training.

The soft spots are real and proportional. No ablation separates image editing from label rescaling. The 1.5× multiplier is asserted, not derived from any braking or time-to-collision model. The improvement over SMOGN on the safety-critical set is 1.6923 vs 1.6966 RMSE—well within plausible noise—and there are no error bars, repeated seeds, or significance tests. The image authenticity claim is supported only by a few figures, not by any quantitative metric. There is also a potential leakage issue: if the augmented frames are added before the top-10% split, some of the 1.5×-relabeled frames themselves fall into the test set. On the citation side, the reference list contains off-topic entries (spacecraft, CT reconstruction, clinical notes), which reads like padding and undermines confidence in the literature engagement.

This is a modest extension of the AR-based augmentation family from reference [17], and the current experiment does not support the central claim. A serious referee would flag the confound immediately. I would not accept this in current form, but the idea is testable and the paper is clearly written. If the authors add a proper ablation (image-only vs label-only vs both), ground the relabeling in kinematics, and report variance, a revised version could be worth another look. For this version, my recommendation is reject.

Referee Report

4 major / 6 minor

Summary. SafeAug proposes to create safety-critical driving examples by detecting the front vehicle in KITTI images, estimating depth with Depth-Anything, reconstructing a 3D point cloud in Open3D, translating the front vehicle closer by half a body length, reprojecting to 2D, and simultaneously multiplying the recorded acceleration label by 1.5. Approximately 200 augmented image-label pairs are added to the original ~2000-image training set. A CNN takes image and speed as input and predicts acceleration. The paper reports that this augmented dataset improves RMSE/MAE on the 10% most safety-critical frames relative to the original dataset, SMOGN, and importance sampling, while roughly maintaining or slightly improving full-dataset performance. The central claim is that the geometric image edit preserves naturalistic authenticity and yields better safety-critical prediction than existing augmentation/sampling baselines.

Significance. The idea of using 3D geometric manipulation of real images to create safety-critical scenes is a plausible alternative to simulation and generative models, and the pipeline is described in a modular way that could be reused. If the reported gains were shown to come from the image edit, the method would be a useful, lightweight augmentation for car-following acceleration prediction. However, in its current form the evidence is confounded and not statistically grounded, so the significance cannot be assessed from the presented experiments. The paper also makes an unsupported authenticity claim. On balance, the concept is interesting but the current validation is insufficient.

major comments (4)

[IV-D, Algorithm 4 step 3; V-C Table I] The improvement on safety-critical RMSE/MAE is confounded: Algorithm 4 changes the input image and rescales the label A_aug = 1.5 × A_orig simultaneously. Because the evaluation set is defined as the 10% of frames with the smallest accelerations (Section V-B), multiplying the labels of the augmented training frames by 1.5 pushes their target values into the very tail where the model is tested. A model trained on those shifted targets will predict more negative accelerations on safety-critical inputs, lowering error on that tail regardless of whether the moved-vehicle image carries any useful visual information. The 1.5 multiplier is asserted in the text ('as the closer vehicle distance at the same speed necessitates a greater deceleration') but no braking or time-to-collision model is provided to tie it to the half-body-length shift. An ablation that trains on (a) image edit with original labels, (b) label rescaling without image edit, and (c) both, is required to attribute the gain to the geometric augmentation.
[V-C, Table I] The comparison lacks error bars, repeated seeds, and significance tests. The safety-critical RMSE difference between Ours (1.6923) and SMOGN (1.6966) is 0.0043, which is two orders of magnitude smaller than the differences between methods on the complete dataset and is likely within run-to-run noise for a small CNN trained on ~2000 images. The authors should report mean and standard deviation over multiple random seeds and a significance test (e.g., paired bootstrap) for both the safety-critical and full-dataset metrics.
[V-B] The dataset-splitting paragraph only defines the safety-critical subset; it does not describe the actual train/test split, the number of training and test frames, the overlap between the ~200 augmented images and the test frames, or whether the augmented versions of test scenes are excluded from training. If the augmented image differs from a test frame only by a half-body-length vehicle shift and the near-duplicate is in the training set, the evaluation may be optimistic. Please specify the split, the selection criteria for the 200 images, and the exclusion rule.
[IV-D, Fig. 4] The claim of 'minimal compromise on image authenticity' is not quantitatively evaluated. The paper provides no metric (e.g., FID, human perceptual study), no comparison of artifact rates, and no ablation measuring the effect of reprojection errors. Since the authenticity of the edited images is the paper's main stated advantage over simulation and generative methods, it should be directly assessed rather than asserted.

minor comments (6)

[Introduction] The phrase 'safety-critical data-anomalous or extreme situations not commonly found in safety-critical data' is confusing and likely contains a typo; please rephrase to distinguish 'safety-critical data' from the general dataset.
[III, Eq. (1)] The notation I_aug(i) ← I(i) is unclear; this appears to be a definition or transformation, so please replace the arrow with a proper mathematical statement or description.
[IV-D, Algorithm 4] The 'Require' block lists I_aug and A_aug as requirements before they are produced; these should be named as outputs or the input/output variables should be separated.
[V-C, Table I] The large degradation of SMOGN on the complete dataset (RMSE 0.4256 vs 0.2416 for Original) is not discussed; this may indicate that SMOGN's synthetic samples harm general performance and should be addressed when interpreting the safety-critical results.
[V-C, Fig. 5] The caption 'Distribution of Acceleration' does not state which datasets or methods are shown or whether the distribution is over training or test data; please clarify.
[References] Several entries (e.g., [3], [4], [8], [9], [14], [15], [20], [21]) appear unrelated to driving data augmentation or safety-critical scenarios; please verify that all references are relevant and cited in context.

Circularity Check

1 steps flagged · score 6.0 of 10

Safety-critical improvement is confounded: Algorithm 4 rescales acceleration labels by 1.5 while also editing the image, so the reported RMSE gain may be produced by relabeling alone; the geometric edit is never isolated.

fitted input called prediction [Algorithm 4, step 3; Section IV-D; Section V-B; Table I]
"Aaug ← 1.5 × Aorig ... we set the acceleration adjustment value to 1.5 times the original acceleration of each image, as the closer vehicle distance at the same speed necessitates a greater deceleration. ... The dataset was sorted according to the vehicle acceleration values and the 10% of data with the smallest acceleration values and the car in front, indicating strong deceleration and potentially hazardous situations, were categorized as safety-critical data."

The augmentation changes two variables at once: the image geometry and the training label (A_aug = 1.5 × A_orig). The safety-critical test set is defined as the 10% of frames with the most negative original accelerations, the same quantity that is multiplied by 1.5 in the augmented training labels. Training on 200 samples with more extreme deceleration labels biases the regressor toward more negative outputs, which directly lowers RMSE/MAE on that tail regardless of whether the moved-vehicle image conveys useful visual information.

full rationale

The paper's central empirical claim -- that the proposed 3D geometric edit of images improves safety-critical acceleration prediction -- is undermined by the simultaneous, un-ablated label rescaling in Algorithm 4. The safety-critical evaluation subset is selected by the most negative original acceleration values, and the augmented training labels are exactly 1.5 times those original accelerations, pushing 200 training targets deeper into the evaluated tail. A regression model trained with such shifted targets will tend to predict more negative accelerations, improving the safety-critical RMSE/MAE by construction, independent of the image edit. The paper provides no physical derivation of the 1.5 factor (no braking model, time-to-collision, or consistency check with the half-body-length distance change) and no ablation separating image augmentation from label augmentation, so the Table I comparison cannot be attributed to the proposed visual transformation. This is a partial circularity/confound rather than a fully definitional equivalence: the 1.5 factor is an asserted input, not a fitted value, but the claimed prediction is statistically tied to that input. No load-bearing self-citation chain or uniqueness-importation issue is present; reference [1] includes an author but is not load-bearing. The paper also lacks repeated seeds/error bars, making small differences (e.g., 1.6923 vs 1.6966) difficult to interpret, but that is a rigor concern, not circularity. Overall score reflects that the main safety-critical improvement may reduce to the label rescaling, warranting a 6.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on four unvalidated premises: (1) Depth-Anything outputs metric depth on KITTI, (2) the front-vehicle selection rule targets the right object, (3) point-cloud reprojection preserves image realism, and (4) scaling acceleration by 1.5 yields physically consistent labels. The first and fourth are the most load-bearing; neither is justified with external measurements or ablations. Several arbitrary constants (1.5 multiplier, half-body distance, 200-image selection, 10% threshold) directly shape the reported improvement.

free parameters (5)

acceleration multiplier = 1.5x
Aaug = 1.5 × Aorig in Algorithm 4, step 3; chosen ad hoc to reflect increased deceleration, no derivation or ablation, and it directly modifies the prediction target.
half body length distance adjustment = unspecified
Equation (6) and Algorithm 4, step 1: the front vehicle is moved closer by 'half body length' of the car; the actual length value and the 3D shift are not specified, and the selection of which point-cloud points belong to the vehicle is not described.
number and selection of augmented images = 200 of ~2000
Section IV-D states images were selected where the initial distance to the front vehicle was already close; the selection criterion is not quantified, and the choice of 200 images is not justified or swept.
safety-critical threshold = 10%
Section V-B2 defines safety-critical data as the 10% of samples with the smallest acceleration values; the threshold is arbitrary and directly affects the reported metric.
CNN architecture and training hyperparameters = unspecified
Section V-B mentions only 'a basic convolutional neural network (CNN) model' with no architecture, optimizer, epochs, learning rate, or random seeds reported, making the downstream result ambiguous.

assumptions (5)

domain assumption Depth-Anything produces metric depth maps suitable for 3D reconstruction.
Section IV-A uses Depth-Anything outputs to build a point cloud; Depth-Anything is typically trained for relative depth, and its absolute scale on KITTI is not validated, so the 3D geometry may be distorted.
domain assumption YOLOv5 detections and the chosen 'front vehicle' rule identify the vehicle whose distance should be modified.
Algorithm 2 picks the largest bounding box intersecting the image center line; this may fail with multiple lanes, curved roads, or partial occlusions, and no validation of detection correctness is provided.
ad hoc to paper Reprojecting the edited point cloud to 2D yields realistic images with no significant artifacts.
Section IV-D asserts 'misaligned shadows are covered' without any quantitative or qualitative evaluation; no user study, FID score, or object-consistency check is reported.
ad hoc to paper KITTI acceleration labels can be multiplied by 1.5 to represent a physically consistent emergency-braking value.
Algorithm 4 sets Aaug = 1.5 × Aorig; no braking dynamics or vehicle model justifies this scaling, and the paper does not test whether the relabeled data is physically plausible.
domain assumption Improvement on the safety-critical subset is the correct measure of augmentation quality.
The test set is the bottom 10% of acceleration values; a model trained on relabeled data may trivially benefit from a shifted output distribution, so the metric may not capture genuine perception improvement.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SafeAug: Safety-Critical Driving Data Augmentation from Naturalistic Datasets." pith.science (2026). https://pith.science/paper/ACLLGEF4

@misc{pith2026250102143,
  author       = {Pith},
  title        = {Pith review of: SafeAug: Safety-Critical Driving Data Augmentation from Naturalistic Datasets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ACLLGEF4}},
  note         = {Machine review of arXiv:2501.02143}
}

read the original abstract

Safety-critical driving data is crucial for developing safe and trustworthy self-driving algorithms. Due to the scarcity of safety-critical data in naturalistic datasets, current approaches primarily utilize simulated or artificially generated images. However, there remains a gap in authenticity between these generated images and naturalistic ones. We propose a novel framework to augment the safety-critical driving data from the naturalistic dataset to address this issue. In this framework, we first detect vehicles using YOLOv5, followed by depth estimation and 3D transformation to simulate vehicle proximity and critical driving scenarios better. This allows for targeted modification of vehicle dynamics data to reflect potentially hazardous situations. Compared to the simulated or artificially generated data, our augmentation methods can generate safety-critical driving data with minimal compromise on image authenticity. Experiments using KITTI datasets demonstrate that a downstream self-driving algorithm trained on this augmented dataset performs superiorly compared to the baselines, which include SMOGN and importance sampling.

Figures

Figures reproduced from arXiv: 2501.02143 by the authors.

**Figure 2.** Street View to Depth [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** YOLO-v5 Result to the camera in 3D space. This adjustment is noticeable and significant because we selected images where the initial distance to the front vehicle was already close. The spatial adjustment is then accurately reflected in the 2D image through a 3D to 2D conversion process. As shown in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Comparison of Vehicle Distance Adjustments [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 5.** Figure 5: Distribution of Acceleration dataset validates the role of the augmented dataset in improving model robustness. Furthermore, [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 22 canonical work pages

[17]

Augmented reality meets computer vision : Efficient data generation for urban driving scenes,

H. A. Alhaija, S. K. Mustikovela, L. Mescheder, A. Geiger, and C. Rother, “Augmented reality meets computer vision : Efficient data generation for urban driving scenes,” 2017

work page 2017
[1]

Uncertainty quantification of collaborative detection for self-driving,

S. Su, Y . Li, S. He, S. Han, C. Feng, C. Ding, and F. Miao, “Uncertainty quantification of collaborative detection for self-driving,” in 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 5588–5594

work page 2023
[2]

Advsim: Generating safety-critical scenarios for self- driving vehicles,

J. Wang, A. Pun, J. Tu, S. Manivasagam, A. Sadat, S. Casas, M. Ren, and R. Urtasun, “Advsim: Generating safety-critical scenarios for self- driving vehicles,” 2023

work page 2023
[3]

Autonomous multi- robot servicing for spacecraft operation extension,

L. Gao, G. Cordova, C. Danielson, and R. Fierro, “Autonomous multi- robot servicing for spacecraft operation extension,” in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2023, pp. 10 729–10 735

2023
[4]

Decentralized Adaptive Aerospace Transportation of Unknown Loads Using A Team of Robots

L. Gao, K. Aubert, D. Saldana, C. Danielson, and R. Fierro, “Decen- tralized adaptive aerospace transportation of unknown loads using a team of robots,” arXiv preprint arXiv:2407.08084 , 2024

work page Pith review arXiv 2024
[5]

Smote: Synthetic minority over-sampling technique,

N. V . Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “Smote: Synthetic minority over-sampling technique,” Journal of Artificial Intelligence Research , vol. 16, p. 321–357, June 2002

work page 2002
[6]

Bayesian estimates of equation system parameters: an application of integration by monte carlo,

T. Kloek and H. K. Van Dijk, “Bayesian estimates of equation system parameters: an application of integration by monte carlo,” Econometrica: Journal of the Econometric Society , pp. 1–19, 1978

work page 1978
[7]

C. X. Ling and V . S. Sheng, Cost-Sensitive Learning. Boston, MA: Springer US, 2010, pp. 231–235

work page 2010

Show all 26 references

[8]

Provably convergent learned inexact descent algorithm for low-dose ct reconstruction,

Q. Zhang, M. Alvandipour, W. Xia, Y . Zhang, X. Ye, and Y . Chen, “Provably convergent learned inexact descent algorithm for low-dose ct reconstruction,” arXiv preprint arXiv:2104.12939 , 2021

2021 arXiv
[9]

Extra proximal-gradient network with learned regularization for image compressive sensing reconstruction,

Q. Zhang, X. Ye, and Y . Chen, “Extra proximal-gradient network with learned regularization for image compressive sensing reconstruction,” Journal of Imaging , vol. 8, no. 7, p. 178, 2022

2022
[10]

Simbar: Single image-based scene relighting for effective data augmentation for automated driving vision tasks,

X. Zhang, N. Tseng, A. Syed, R. Bhasin, and N. Jaipuria, “Simbar: Single image-based scene relighting for effective data augmentation for automated driving vision tasks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp...

2022
[11]

Surfelgan: Synthesizing realistic sensor data for autonomous driving,

Z. Yang, Y . Chai, D. Anguelov, Y . Zhou, P. Sun, D. Erhan, S. Rafferty, and H. Kretzschmar, “Surfelgan: Synthesizing realistic sensor data for autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2020

2020
[12]

King: Generating safety-critical driving scenarios for robust imitation via kinematics gradients,

N. Hanselmann, K. Renz, K. Chitta, A. Bhattacharyya, and A. Geiger, “King: Generating safety-critical driving scenarios for robust imitation via kinematics gradients,” 2022

2022
[13]

Adv3d: Generating safety-critical 3d objects through closed-loop simulation,

J. Sarva, J. Wang, J. Tu, Y . Xiong, S. Manivasagam, and R. Urtasun, “Adv3d: Generating safety-critical 3d objects through closed-loop simulation,” 2023

2023
[14]

Real-time system optimal traffic routing under uncertainties–can physics models boost reinforcement learning?

Z. Ke, Q. Zou, J. Liu, and S. Qian, “Real-time system optimal traffic routing under uncertainties–can physics models boost reinforcement learning?” arXiv preprint arXiv:2407.07364 , 2024

2024 arXiv
[15]

Safety effects of road pavement resurfacing: A case study of city-wide scale projects in china,

P. Ouyang, X. Li, C. Xu, J. Wu, and Z. Ke, “Safety effects of road pavement resurfacing: A case study of city-wide scale projects in china,” Journal of Transportation Safety & Security , vol. 14, no. 3, pp. 453–472, 2022

2022
[16]

Occgan: Semantic image augmentation for driving scenes,

Y . Wang, L. Mo, H. Ma, and J. Yuan, “Occgan: Semantic image augmentation for driving scenes,” Pattern Recognition Letters , vol. 136, pp. 257–263, 2020

2020
[18]

Radar: Adversarial driving style representation learning with data augmentation,

Z. Liu, J. Zheng, J. Lin, L. Wang, and K. Wu, “ Radar: Adversarial driving style representation learning with data augmentation,” IEEE Transactions on Mobile Computing , vol. 22, no. 12, pp. 7070–7085, 2023

2023
[19]

Tauaud: Test augmentation of image recognition in autonomous driving,

S. Zhang, J. Liu, B. Xu, and G. Liu, “Tauaud: Test augmentation of image recognition in autonomous driving,” in 2021 IEEE 21st International Conference on Software Quality, Reliability and Security Companion (QRS-C), 2021, pp. 550–554

2021
[20]

Fineehr: Refine clinical note repre- sentations to improve mortality prediction,

J. Wu, X. Ye, C. Mou, and W. Dai, “Fineehr: Refine clinical note repre- sentations to improve mortality prediction,” in 2023 11th International Symposium on Digital Forensics and Security (ISDFS) . IEEE, 2023, pp. 1–6

2023
[21]

Medlens: Improve mortality prediction via medical signs selecting and regression interpolation,

X. Ye, J. Wu, C. Mou, and W. Dai, “Medlens: Improve mortality prediction via medical signs selecting and regression interpolation,” arXiv preprint arXiv:2305.11742 , 2023

2023 arXiv
[22]

Depth anything: Unleashing the power of large-scale unlabeled data,

L. Yang, B. Kang, Z. Huang, X. Xu, J. Feng, and H. Zhao, “Depth anything: Unleashing the power of large-scale unlabeled data,” in CVPR, 2024

2024
[23]

Yolov5 by ultralytics,

G. Jocher, “Yolov5 by ultralytics,” 2020

2020
[24]

Open3D: A modern library for 3D data processing,

Q.-Y . Zhou, J. Park, and V . Koltun, “Open3D: A modern library for 3D data processing,” arXiv:1801.09847, 2018

2018 arXiv
[25]

Vision meets robotics: The kitti dataset,

A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,” International Journal of Robotics Research (IJRR) , 2013

2013
[26]

SMOGN: Synthetic minority over-sampling technique for regression with gaussian noise,

N. Kunz, “SMOGN: Synthetic minority over-sampling technique for regression with gaussian noise,” 2020

2020

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [17]

Augmented reality meets computer vision : Efficient data generation for urban driving scenes,

H. A. Alhaija, S. K. Mustikovela, L. Mescheder, A. Geiger, and C. Rother, “Augmented reality meets computer vision : Efficient data generation for urban driving scenes,” 2017

work page 2017

[2] [1]

Uncertainty quantification of collaborative detection for self-driving,

S. Su, Y . Li, S. He, S. Han, C. Feng, C. Ding, and F. Miao, “Uncertainty quantification of collaborative detection for self-driving,” in 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 5588–5594

work page 2023

[3] [2]

Advsim: Generating safety-critical scenarios for self- driving vehicles,

J. Wang, A. Pun, J. Tu, S. Manivasagam, A. Sadat, S. Casas, M. Ren, and R. Urtasun, “Advsim: Generating safety-critical scenarios for self- driving vehicles,” 2023

work page 2023

[4] [3]

Autonomous multi- robot servicing for spacecraft operation extension,

L. Gao, G. Cordova, C. Danielson, and R. Fierro, “Autonomous multi- robot servicing for spacecraft operation extension,” in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2023, pp. 10 729–10 735

2023

[5] [4]

Decentralized Adaptive Aerospace Transportation of Unknown Loads Using A Team of Robots

L. Gao, K. Aubert, D. Saldana, C. Danielson, and R. Fierro, “Decen- tralized adaptive aerospace transportation of unknown loads using a team of robots,” arXiv preprint arXiv:2407.08084 , 2024

work page Pith review arXiv 2024

[6] [5]

Smote: Synthetic minority over-sampling technique,

N. V . Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “Smote: Synthetic minority over-sampling technique,” Journal of Artificial Intelligence Research , vol. 16, p. 321–357, June 2002

work page 2002

[7] [6]

Bayesian estimates of equation system parameters: an application of integration by monte carlo,

T. Kloek and H. K. Van Dijk, “Bayesian estimates of equation system parameters: an application of integration by monte carlo,” Econometrica: Journal of the Econometric Society , pp. 1–19, 1978

work page 1978

[8] [7]

C. X. Ling and V . S. Sheng, Cost-Sensitive Learning. Boston, MA: Springer US, 2010, pp. 231–235

work page 2010