Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring

Lei He; Ludan Zhang; Qinyue Luo; Shuofei Qiao; Sihan Wang; Yuqi Dai

REVIEW 5 major objections 6 minor 21 references

Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring

T0 review · 5 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a Feature Map Quality Score, computed from NDS ratios and cosine similarity to a top checkpoint and predicted by a CLIP-based network, can serve as an auxiliary loss that improves BEVFormer 3D detection on nuScenes by

desk verdict The paper's central claim is confounded: the auxiliary loss is 80% a normalized NDS, so the 3.89% gain isn't evidence for the feature-quality score; still worth a referee. read the letter →

arxiv 2508.07552 v2 pith:YV2ZKK2R submitted 2025-08-11 cs.CV

Ludan Zhang , Sihan Wang , Yuqi Dai , Shuofei Qiao , Qinyue Luo , Lei He This is my paper · ORCID

classification cs.CV

keywords featuremapqualityscoringfunctionalmoduleevaluationautonomousdrivingperception3DobjectdetectionBEVFormerCLIPauxiliarylossnuScenes

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

End-to-end driving models lack explicit supervision for their internal modules, so developers cannot tell how mature a module is or guide it directly. This paper proposes a Feature Map Quality Score (FMQS) that combines a macro-level NDS ratio—how close a module configuration's detection score is to the best of 64 configuration-stage runs—with a micro-level channel-spatial cosine similarity between the module's feature map and that best checkpoint's feature map. It then trains CLIP-FMQE-Net, a lightweight network that aligns feature maps with CLIP text embeddings of ground-truth annotations and regresses FMQS, and uses the predicted score as an auxiliary loss during BEVFormer training. On nuScenes, this yields an average 3.89% NDS gain in 3D object detection, with the image and BEV feature modules helping most when applied together. If correct, the result means hidden module quality can be quantified and optimized without task-level labels for the intermediate representations.

What carries the argument

The object that carries the argument is the Feature Map Quality Score (FMQS), computed as the weighted fusion (with $w=0.8$) of a macro-level score—the current configuration-stage NDS divided by the SOTA NDS—and a micro-level score, the Channel-Spatial Cosine Similarity between the module's feature map and the SOTA checkpoint's feature map, averaging cosine similarity over channels and spatial positions. Around this scalar, the paper builds a dual-granularity scoring system that generates training labels for stored feature maps, a CLIP-based encoder pair that projects feature maps and textualized ground truth into one semantic space, and a Transformer-decoder head that predicts the score. Th

What would settle it

Run BEVFormer training with the FMQS auxiliary loss replaced by (a) a constant, (b) a random score, or (c) the macro-level NDS ratio alone. If any of these reproduce the 3.89% NDS gain, the micro-level feature-similarity signal is not what carries the improvement. Separately, freeze a module's weights after random initialization, or feed shuffled input, and check that its FMQS drops below the SOTA's; a single case where a degraded or random feature map scores at or above the SOTA map falsifies the quality interpretation.

Watch

Extended reading notes

Core claim

The central claim is that a functional module's training maturity can be captured by a scalar score, FMQS, built as a weighted fusion (with weight $w=0.8$) of a task-level score—the module's NDS divided by the NDS of the best configuration-stage checkpoint—and a structure-level score, the Channel-Spatial Cosine Similarity between the module's feature map and that checkpoint's feature map. The paper further claims that this score is learnable: CLIP-FMQE-Net, using a CLIP text encoder for ground truth and a light CNN for feature maps, aligns the two modalities in a shared semantic space and regresses FMQS with a Transformer decoder, reaching average $R^2 = 0.855$ for the image feature extracti

Load-bearing premise

The pipeline depends on the premise that a feature map's quality is faithfully measured by its channel-spatial cosine similarity to one chosen best checkpoint's feature map, supplemented by an NDS ratio, and that CLIP text embeddings of annotations share a semantic space with BEV feature maps; if either premise fails, the score, its regression labels, and the auxiliary loss lose their meaning.

Editorial extensions

If this is right

Developers can monitor module maturity during training from a single predicted score, without waiting for final task metrics.
The auxiliary loss can be applied during BEVFormer training with frozen CLIP-FMQE-Net parameters, so the evaluator does not need to be retrained for each run.
Jointly supervising both the image-level and BEV-level feature modules gives a larger average gain (3.89% NDS) than either alone, suggesting multi-module quality signals compound.
Because FMQS correlates with final NDS, it offers a cheap selection criterion for configuration and checkpoint choice in multi-module learning pipelines.
The approach transfers CLIP's cross-modal alignment to a regression setting where only a few hundred feature-map samples are needed to predict quality.

Reading between the lines

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

The method's reference is a single SOTA checkpoint; if that checkpoint is unrepresentative, the score inherits its biases. Replacing it with an ensemble or a task-derived ideal feature distribution is a natural stress test the paper does not run.
The CLIP text encoder may be doing less semantic work than claimed: because ground-truth text is template-based, a fixed learned embedding of annotation classes might align feature maps just as well, at lower cost.
The 3.89% average gain mixes large gains (up to 7.36%) with much smaller ones (e.g. 0.70% in one configuration), so the mechanism's benefit may depend on which backbone and head are used.
The same FMQS formulation would apply to other tensor-producing modules—for example, occupancy or trajectory predictors—provided their ground truth can be rendered as text templates; testing that extension would clarify whether the score is generic or BEV-specific.

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 headline is that this is a reasonable extension of the authors' earlier RSS work [19], but the main claimed result—that the FMQS auxiliary loss improves BEVFormer—is confounded by construction. The reported 3.89% NDS gain should not be read as evidence that the feature-quality score works as advertised.

The genuinely new piece is converting the evaluation score into a trainable auxiliary loss, using CLIP to align feature maps with text-encoded ground truth. That is a plausible direction, and the regression pipeline is described clearly enough to reproduce the FMQS labels. The regression numbers (R2 ~0.85 for IFEM) look respectable, though they don't tell you much once you notice the label structure.

The load-bearing problem is Eqs. (6) and (12). FMQS is 80% normalized NDS and 20% CS-CosSim. The auxiliary loss is 1 minus the predicted FMQS. Since the predictor is trained to track the true FMQS, the gradient signal is dominated by a monotone function of the target metric. So the loss would push NDS up even if the CLIP alignment and similarity term were pure noise. The paper does not ablate the NDS component, which is the one control that would separate the quality-score hypothesis from the trivial "optimize a proxy for NDS" explanation. The R2 values are also inflated by the fact that NDS/NDS_sota is a stage-level constant shared by all samples in a configuration; the model can learn the maturity curve without measuring per-sample feature-map quality.

The integration results are weaker than presented. For several configurations, the combined IFEM+BFEM gain is below the individual gains (e.g., Res50-SCA-TSA: 5.72% and 5.74% individually, 4.46% combined; VoV-SCA-RCF: 1.58/1.63 vs 1.32). The text says the joint curve is "consistently" better, but Table 4 contradicts that. With no baseline NDS values, only relative percentages, and no variance across seeds, the average 3.89% is not established.

Minor issues: the loss weights w_BEV, w_IFEM, w_BFEM and the temperature tau are not reported. A comparison to a simple auxiliary loss that directly uses NDS, or to a standard feature-distillation baseline, is absent. That would be the real test.

To be fair, the paper is a direct and honest extension of the authors' prior work, and the writing is mostly clear. The core idea of using a learned feature-quality score as supervision is worth exploring. But as presented, the evidence doesn't separate the interesting part from the trivial NDS proxy.

I would not cite this in its current form, but I would send it to peer review rather than desk-reject: the idea is significant enough, and the flaw is fixable with the right control experiments. A referee can ask for an NDS-ablated loss, absolute metric values, error bars, and the missing hyperparameters.

Referee Report

5 major / 6 minor

Summary. The paper proposes a Feature Map Quality Score (FMQS) for evaluating intermediate functional modules in BEVFormer-style autonomous driving models. FMQS combines a macro-level model-performance score (normalized NDS) with a micro-level feature-map similarity score (CS-CosSim against a SOTA checkpoint), weighted 0.8/0.2. A CLIP-based network, CLIP-FMQE-Net, is trained to regress FMQS from feature maps and ground-truth text, and the predicted score is then used as an auxiliary loss during BEVFormer training. Experiments on nuScenes report high regression accuracy (average R² ≈ 0.855 for IFEM, 0.794 for BFEM) and an average 3.89% NDS improvement when the auxiliary loss is integrated. The central claim is that FMQS provides an independent, interpretable quality signal that improves detection performance.

Significance. If validated, the idea of a decoupled, quantitative quality score for intermediate feature maps would be useful for modular perception systems: it could enable independent module monitoring, configuration selection, and auxiliary supervision without changing the task head. The paper also addresses a real gap in the BEV perception literature, where intermediate feature quality is rarely evaluated separately from final task metrics. The authors provide a concrete scoring formula, a regression network, and a training-integration recipe. However, the paper currently lacks the experimental controls needed to establish that the proposed quality signal, rather than the NDS component embedded in the score, drives the reported gains. The significance is therefore conditional on a component-level ablation that the manuscript does not provide.

major comments (5)

[§3.2.3, Eqs. (2), (6), (12)] The auxiliary loss is not independent of the target metric. Eq. (6) defines FMQS = 0.8·Score_model + 0.2·Score_feature, and Eq. (2) defines Score_model = NDS/NDS_sota. The training loss in Eq. (12) is L_FMQS = 1 − FMQS_pred. Even if the predictor were perfect, 80% of the training signal is a monotone transform of NDS, so the reported NDS gain is expected even if both the CS-CosSim term and the CLIP alignment contribute nothing. The paper needs an ablation that isolates the 0.2 similarity component, e.g., comparing against a loss using only Score_model, only Score_feature, and a directly injected NDS-based reward. Without this, the central claim that the learned feature-quality score improves detection is not established.
[§4.4, Table 4] The text claims that joint integration 'consistently achieves higher NDS than single-module integration,' but Table 4 contradicts this for three of eight configurations: Res50-SCA-TSA has individual gains 5.72% and 5.74% but a combined gain of 4.46%; VoV-SCA-RCF has 1.58%/1.63% individually vs 1.32% combined; VoV-GKT-TSA has 2.24%/2.36% vs 0.70% combined. Moreover, the table reports no baseline NDS values, no standard deviations, and no number of seeds, so the average 3.89% gain is not statistically supported. At minimum, the paper should report mean ± std over multiple runs and provide the baseline NDS for each configuration.
[§4.1–§4.2, Eqs. (1)–(6)] The regression labels are dominated by a stage-level constant. For a given configuration–stage pair, Score_model is identical for all samples, and it contributes 80% of the FMQS label. The per-sample variation comes only from the 0.2-weighted CS-CosSim term. The reported R² ≈ 0.85/0.79 may therefore reflect the predictor learning configuration/stage identity rather than per-sample feature-map quality. The paper should report the R² after removing the stage-level component (e.g., predicting only the CS-CosSim residual) or compare against a baseline that uses one-hot configuration/stage features.
[§4.4, Eq. (13)] The loss weights w_BEV, w_IFEM, and w_BFEM in Eq. (13) are never reported, nor is the relative scale of the FMQS losses to the BEV detection loss. Because the auxiliary loss includes a large NDS-normalized term, the reported improvements could be an artifact of weighting rather than of the proposed quality scoring. The paper should report these weights, the gradient scales, and a sensitivity analysis over at least a small range of weights.
[§4.2 (FMQS-Dataset) and §4.1 (Table 1)] The FMQS regression dataset is built from nuScenes-mini (404 keyframe samples), while the NDS values in Table 1 are presumably computed on the full nuScenes validation set. The paper does not clarify whether the NDS used in Eq. (2) is computed on the same mini split used to store feature maps. If the NDS comes from the full validation set, then the macro-level score does not correspond to the stored mini-split feature maps, undermining the label construction. Please specify the evaluation protocol for the NDS values used in FMQS.

minor comments (6)

[§3.2.2, Eq. (2)] Notation issue: 'Score isota,jsota model' should be formatted as Score_{i_sota, j_sota}^{model}; also 'isota' appears without proper subscripting.
[§4.2] Typo: 'activvation' should be 'activation'.
[Table 4] The header 'IEFM' should read 'IFEM' for consistency with the Image Feature Extraction Module.
[§3.2] Typo: 'constructe' should be 'construct'.
[§3.3, Eq. (9)] The temperature τ in the contrastive loss is introduced but its value is never reported. Please provide the setting used in the experiments.
[General] No code or trained model weights are provided. Given that the method relies on a custom scoring pipeline and the nuScenes-mini regression dataset, releasing code would substantially aid reproducibility.

Circularity Check

1 steps flagged · score 8.0 of 10

FMQS is 80% normalized NDS by Eq. (6), so the auxiliary loss in Eq. (12) directly optimizes the reported evaluation metric; the NDS gain is built into the loss definition.

self definitional [Section 3.2.2 Eq. (2), Section 3.2.3 Eq. (6), Section 3.4 Eq. (12)]
"Score i,j model = N DSi,j / N DSsota . (2) ... F M QSi,j,k = w · Score i,j model + (1 − w) · Score i,j,k F eature, (6) Here, the weighting coefficient w is set to 0.8. ... LIF EM FMQS= 1 − ^F M QSIF EM, LBF EM FMQS= 1 − ^F M QSBF EM, (12)"

By Eq. (2), the macro-level component of FMQS is exactly the target metric NDS normalized by the SOTA NDS. By Eq. (6), 80% of FMQS is this NDS ratio. By Eq. (12), the auxiliary loss is 1 minus the predicted FMQS. Therefore, up to the regression error and the 20% similarity term, the auxiliary loss is 0.8·(1 − NDS/NDS_SOTA) + 0.2·(1 − CS-CosSim). Minimizing it directly maximizes a monotone transform of the very metric (NDS) used to report the 3.89% gain. Thus the reported improvement is expected even if the feature-similarity and CLIP components contribute nothing. The regression labels are likewise NDS-dominated, so high R2 mainly reflects predicting a configuration/stage-level constant rather than per-sample feature-map quality. The claim that FMQS measures feature-map quality and improve

full rationale

The central circularity is definitional, not a matter of author intent. Eq. (6) defines FMQS as 0.8·(NDS/NDS_SOTA) + 0.2·CS-CosSim, and Eq. (12) converts 1 − FMQS into the auxiliary training loss. Since 80% of the auxiliary loss is a decreasing function of NDS, adding this loss to BEVFormer should raise NDS even if the feature-map-similarity and CLIP-alignment machinery are wholly uninformative. The paper's headline result—an average 3.89% NDS gain—is therefore not independent evidence that the learned score captures feature-map quality; it is partly a consequence of optimizing a loss that contains the evaluation metric itself. The regression results (R2 ≈ 0.855/0.794) are also consistent with predicting the NDS-ratio component, which is constant across samples within each configuration-stage cell, rather than per-sample quality. Self-citations to [18] and [19] are present and the framework is inherited from the authors' prior work, but the decisive reduction is visible in the paper's own equations, so the score is driven by the definitional circularity rather than by self-citation alone. Other issues, such as the inconsistency in Table 4 where combined integration sometimes underperforms individual modules (e.g., 5.72%/5.74% individually vs 4.46% combined for Res50-SCA-TSA; 1.58%/1.63% vs 1.32% for VoV-SCA-RCF), are correctness concerns rather than circularity and do not affect this score.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method depends on a hand-chosen SOTA reference, two fixed weights (alpha=0.5, w=0.8), unreported loss weights and temperature, and three domain assumptions about feature similarity, NDS sufficiency, and CLIP alignment. None are independently validated.

free parameters (5)

alpha (channel-spatial balance) = 0.5
Eq. 3; set to 0.5 with no sensitivity study, controls CS-CosSim.
w (macro/micro fusion weight) = 0.8
Eq. 6; chosen by hand, makes FMQS 80% NDS-based.
temperature tau in contrastive loss = not reported
Eq. 9; required to train CLIP-FMQE-Net, value omitted.
loss weights w_BEV, w_IFEM, w_BFEM = not reported
Eq. 13; essential to total loss, values omitted, so training cannot be reproduced.
SOTA reference checkpoint = Config 6, Stage 7 (VoV-SCA-RCF)
Section 4.1; selected as argmax NDS from the authors' own runs; all scores and gains are relative to this choice and no external SOTA is used.

assumptions (4)

domain assumption Feature map quality is measured by similarity to the feature maps of a single SOTA checkpoint (CS-CosSim).
Section 3.2.3, Eqs. 3-5; no external validation that SOTA similarity tracks true quality.
domain assumption NDS is a sufficient criterion for module training maturity and for selecting the SOTA reference.
Sections 3.2.1 and 3.2.2; NDS is a whole-system metric, not a per-module criterion, and it is folded into FMQS itself.
domain assumption CLIP text embeddings of structured annotation text are semantically aligned with BEV feature maps after contrastive training.
Section 3.3; the paper assumes cross-modal alignment transfers from natural images/text to BEV camera feature maps and nuScenes label text; not demonstrated.
domain assumption The FMQS regression head trained on 323 nuScenes-mini samples generalizes to the training distribution and to new configurations.
Sections 4.2-4.4; no out-of-configuration test; small data set.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring." pith.science (2026). https://pith.science/paper/YV2ZKK2R

@misc{pith2026250807552,
  author       = {Pith},
  title        = {Pith review of: Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YV2ZKK2R}},
  note         = {Machine review of arXiv:2508.07552}
}

read the original abstract

End-to-end models are emerging as the mainstream in autonomous driving perception and planning. However, the lack of explicit supervision signals for intermediate functional modules leads to opaque operational mechanisms and limited interpretability, making it challenging for traditional methods to independently evaluate and train these modules. Pioneering in the issue, this study builds upon the feature map-truth representation similarity-based evaluation framework and proposes an independent evaluation method based on Feature Map Convergence Score (FMCS). A Dual-Granularity Dynamic Weighted Scoring System (DG-DWSS) is constructed, formulating a unified quantitative metric - Feature Map Quality Score - to enable comprehensive evaluation of the quality of feature maps generated by functional modules. A CLIP-based Feature Map Quality Evaluation Network (CLIP-FMQE-Net) is further developed, combining feature-truth encoders and quality score prediction heads to enable real-time quality analysis of feature maps generated by functional modules. Experimental results on the NuScenes dataset demonstrate that integrating our evaluation module into the training improves 3D object detection performance, achieving a 3.89 percent gain in NDS. These results verify the effectiveness of our method in enhancing feature representation quality and overall model performance.

Figures

Figures reproduced from arXiv: 2508.07552 by the authors.

**Figure 1.** Independent Evaluation Method for Complex Modules Based on FMQS. CLIP-FMQE-Net enables quality assessment of feature maps by aligning them with structured ground truth in a shared semantic space and predicting scores via Transformer-based decoding. the proposed evaluation module is effectively integrated into the training process of 3D object detection tasks, allowing for adaptive guidance and continuous optimizatio… view at source ↗

**Figure 2.** Feature Map Quality Score (FMQS) Generation Process. Macro-level scores are computed via NDS normalization across configurations, while micro-level scores measure feature similarity to the SOTA model using CS-CosSim. gorized into three types. The first category leverages statistical metrics, such as mutual information [10] and structural similarity [16], to assess spatial consistency of feature maps before and afte… view at source ↗

**Figure 3.** Training Results of BEVFormer with Integrated Independent Evaluation Modules for 3D Obstacle Detection [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 17 canonical work pages

[19]

Unveiling the Black Box: Independent Functional Module Evaluation for Bird's-Eye-View Perception Model

Ludan Zhang, Xiaokang Ding, Yuqi Dai, Lei He, and Keqiang Li. Unveiling the black box: Independent func- tional module evaluation for bird’s-eye-view perception model. arXiv preprint arXiv:2409.11969, 2024. 9

work page Pith review arXiv 2024
[1]

Neural module networks

Jacob Andreas, Marcus Rohrbach, Trevor Darrell, and Dan Klein. Neural module networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 39–48, 2016

2016
[2]

nuscenes: A multimodal dataset for autonomous driving

Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A multimodal dataset for autonomous driving. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020

work page 2020
[3]

Grad-cam++: Gener- alized gradient-based visual explanations for deep con- volutional networks

Aditya Chattopadhay, Anirban Sarkar, Prantik Howlader, and Vineeth N Balasubramanian. Grad-cam++: Gener- alized gradient-based visual explanations for deep con- volutional networks. In 2018 IEEE winter conference on applications of computer vision (WACV), pages 839–847. IEEE, 2018

work page 2018
[4]

Hierarchical and Decoupled BEV Perception Learning Framework for Autonomous Driving

Yuqi Dai, Jian Sun, Shengbo Eben Li, Qing Xu, Jianqiang Wang, Lei He, and Keqiang Li. Hierarchical and decou- pled bev perception learning framework for autonomous driving. arXiv preprint arXiv:2407.12491, 2024

work page Pith review arXiv 2024
[5]

Clip-rldrive: Human-aligned autonomous driving via clip-based re- ward shaping in reinforcement learning

Erfan Doroudian and Hamid Taghavifar. Clip-rldrive: Human-aligned autonomous driving via clip-based re- ward shaping in reinforcement learning. arXiv preprint arXiv:2412.16201, 2024

arXiv 2024
[6]

Pathnet: Evolution channels gra- dient descent in super neural networks

Chrisantha Fernando, Dylan Banarse, Charles Blundell, Yori Zwols, David Ha, Andrei A Rusu, Alexander Pritzel, and Daan Wierstra. Pathnet: Evolution channels gra- dient descent in super neural networks. arXiv preprint arXiv:1701.08734, 2017

arXiv 2017
[7]

BEV-TSR: Text-Scene Retrieval in BEV Space for Autonomous Driving

Zhengyu Jia, Tian Gao, Changwei Cai, Chengkai Hou, Peng Jia, Fan JingChen, YIXING ZHAO, Kun Zhan, FU LIU, YANG WANG, et al. Bev-clip: Multi-modal bev retrieval methodology for complex scene in autonomous driving. arXiv preprint arXiv:2401.01065v1, 2024

work page Pith review arXiv 2024

Show all 21 references

[8]

Bevdepth: Acquisition of reliable depth for multi-view 3d object detection

Yinhao Li, Zheng Ge, Guanyi Yu, Jinrong Yang, Zen- gran Wang, Yukang Shi, Jianjian Sun, and Zeming Li. Bevdepth: Acquisition of reliable depth for multi-view 3d object detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 1477–1485, 2023

2023
[9]

Bev- former: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers

Zhiqi Li, Wenhai Wang, Hongyang Li, Enze Xie, Chong- hao Sima, Tong Lu, Qiao Yu, and Jifeng Dai. Bev- former: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2024

2024
[10]

Multimodality im- age registration by maximization of mutual information

Frederik Maes, Andre Collignon, Dirk Vandermeulen, Guy Marchal, and Paul Suetens. Multimodality im- age registration by maximization of mutual information. IEEE transactions on Medical Imaging , 16(2):187–198, 2002

2002
[11]

Supervised quality assess- ment of medical image registration: Application to intra- patient ct lung registration

Sascha EA Muenzing, Bram van Ginneken, Keelin Mur- phy, and Josien PW Pluim. Supervised quality assess- ment of medical image registration: Application to intra- patient ct lung registration. Medical image analysis , 16(8):1521–1531, 2012

2012
[12]

Clip-bevformer: Enhancing multi-view image-based bev detector with ground truth flow

Chenbin Pan, Burhaneddin Yaman, Senem Velipasalar, and Liu Ren. Clip-bevformer: Enhancing multi-view image-based bev detector with ground truth flow. InPro- ceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 15216–15225, 2024

2024
[13]

Learning transferable visual models from natural lan- guage supervision

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural lan- guage supervision. In International conference on ma- chine learni...

2021
[14]

” why should i trust you?” explaining the predictions of any classifier

Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. ” why should i trust you?” explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1135–1144, 2016

2016
[15]

Shapley explanation networks

Rui Wang, Xiaoqian Wang, and David I Inouye. Shapley explanation networks. arXiv preprint arXiv:2104.02297, 2021

2021 arXiv
[16]

Image quality assessment: Form error visibil- ity to structural similarity

Z Wang. Image quality assessment: Form error visibil- ity to structural similarity. IEEE Trans. Image Process., 13(4):604–606, 2004

2004
[17]

Bevformer v2: Adapt- ing modern image backbones to bird’s-eye-view recogni- tion via perspective supervision

Chenyu Yang, Yuntao Chen, Hao Tian, Chenxin Tao, Xizhou Zhu, Zhaoxiang Zhang, Gao Huang, Hongyang Li, Yu Qiao, Lewei Lu, et al. Bevformer v2: Adapt- ing modern image backbones to bird’s-eye-view recogni- tion via perspective supervision. In Proceedings of the IEEE/CVF conferen...

2023
[18]

Feature map convergence evaluation for functional mod- ule

Ludan Zhang, Chaoyi Chen, Lei He, and Keqiang Li. Feature map convergence evaluation for functional mod- ule. In 2024 IEEE 27th International Conference on In- telligent Transportation Systems (ITSC), pages 864–869. IEEE, 2024

2024
[20]

Anomaly detection of structural health monitoring data using the maximum likelihood estimation-based bayesian dynamic linear model

Yi-Ming Zhang, Hao Wang, Hua-Ping Wan, Jian-Xiao Mao, and Yi-Chao Xu. Anomaly detection of structural health monitoring data using the maximum likelihood estimation-based bayesian dynamic linear model. Struc- tural Health Monitoring, 20(6):2936–2952, 2021

2021
[21]

Deformable detr: Deformable transformers for end-to-end object detection

Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable detr: Deformable transformers for end-to-end object detection. arXiv preprint arXiv:2010.04159, 2020

2010 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [19]

Unveiling the Black Box: Independent Functional Module Evaluation for Bird's-Eye-View Perception Model

Ludan Zhang, Xiaokang Ding, Yuqi Dai, Lei He, and Keqiang Li. Unveiling the black box: Independent func- tional module evaluation for bird’s-eye-view perception model. arXiv preprint arXiv:2409.11969, 2024. 9

work page Pith review arXiv 2024

[2] [1]

Neural module networks

Jacob Andreas, Marcus Rohrbach, Trevor Darrell, and Dan Klein. Neural module networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 39–48, 2016

2016

[3] [2]

nuscenes: A multimodal dataset for autonomous driving

Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A multimodal dataset for autonomous driving. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020

work page 2020

[4] [3]

Grad-cam++: Gener- alized gradient-based visual explanations for deep con- volutional networks

Aditya Chattopadhay, Anirban Sarkar, Prantik Howlader, and Vineeth N Balasubramanian. Grad-cam++: Gener- alized gradient-based visual explanations for deep con- volutional networks. In 2018 IEEE winter conference on applications of computer vision (WACV), pages 839–847. IEEE, 2018

work page 2018

[5] [4]

Hierarchical and Decoupled BEV Perception Learning Framework for Autonomous Driving

Yuqi Dai, Jian Sun, Shengbo Eben Li, Qing Xu, Jianqiang Wang, Lei He, and Keqiang Li. Hierarchical and decou- pled bev perception learning framework for autonomous driving. arXiv preprint arXiv:2407.12491, 2024

work page Pith review arXiv 2024

[6] [5]

Clip-rldrive: Human-aligned autonomous driving via clip-based re- ward shaping in reinforcement learning

Erfan Doroudian and Hamid Taghavifar. Clip-rldrive: Human-aligned autonomous driving via clip-based re- ward shaping in reinforcement learning. arXiv preprint arXiv:2412.16201, 2024

arXiv 2024

[7] [6]

Pathnet: Evolution channels gra- dient descent in super neural networks

Chrisantha Fernando, Dylan Banarse, Charles Blundell, Yori Zwols, David Ha, Andrei A Rusu, Alexander Pritzel, and Daan Wierstra. Pathnet: Evolution channels gra- dient descent in super neural networks. arXiv preprint arXiv:1701.08734, 2017

arXiv 2017

[8] [7]

BEV-TSR: Text-Scene Retrieval in BEV Space for Autonomous Driving

Zhengyu Jia, Tian Gao, Changwei Cai, Chengkai Hou, Peng Jia, Fan JingChen, YIXING ZHAO, Kun Zhan, FU LIU, YANG WANG, et al. Bev-clip: Multi-modal bev retrieval methodology for complex scene in autonomous driving. arXiv preprint arXiv:2401.01065v1, 2024

work page Pith review arXiv 2024