pith. sign in

arxiv: 2606.31834 · v1 · pith:6O5UUUBZnew · submitted 2026-06-30 · 💻 cs.CV · cs.AI

Real-Time Source-Free Object Detection

Sairam VCR , Varun Gopal , Poornima Jain , Vineeth N Balasubramanian , Muhammad Haris Khan This is my paper

Pith reviewed 2026-07-01 05:53 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords source-free object detectiondomain adaptationreal-time detectiondual-head detectormean-teacherpseudo-label fusionmulti-scale features
0
0 comments X

The pith

Dual-head detectors with selective pseudo-label fusion and multi-scale feature diversification enable faster, more compact source-free object detection.

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

The paper aims to show that source-free object detection can meet real-time constraints by building on NMS-free dual-head detectors such as YOLOv10. It finds that standard mean-teacher self-training produces poor results on these detectors because simple pseudo-label strategies from one or both heads give weak supervision and because domain shift reduces the discriminability of multi-scale features. DHF selectively combines one-to-one and one-to-many predictions from the two heads to improve label quality, while MARD adds variance and covariance constraints to keep multi-scale features informative. Both changes operate only at training time. The resulting RT-SFOD method reports higher accuracy than prior SFOD approaches together with higher throughput and roughly half the parameters.

Core claim

Directly applying vanilla mean-teacher to dual-head detectors yields suboptimal adaptation because single-head or naive combination pseudo-labels are imprecise under domain shift and because multi-scale feature maps lose discriminability. DHF selectively admits one-to-one and one-to-many head predictions to preserve precision while recovering missed objects. MARD enforces detection-aware variance and covariance constraints on multi-scale feature maps to restore discriminability. Both modules are training-only and leave inference unchanged, producing 1.4 to 3.5 percent mAP gains, 1.3 times higher throughput, and about two times fewer parameters than prior state-of-the-art SFOD methods.

What carries the argument

DHF (Dual-Head Pseudo-Label Fusion) that selectively admits O2O and O2M predictions from the two heads, together with MARD (Multi-scale Adaptive Representation Diversification) loss that applies variance and covariance constraints on multi-scale feature maps.

If this is right

  • Yields 1.4 to 3.5 percent mAP gains across domain-shift benchmarks.
  • Achieves 1.3 times higher throughput than prior SFOD methods.
  • Uses roughly two times fewer parameters while maintaining or improving accuracy.
  • Generalizes to additional YOLO- and DETR-based dual-head detectors.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The training-only nature of the modules suggests they can be combined with other adaptation techniques without changing deployed model size or latency.
  • The same fusion and diversification ideas could be tested on other self-supervised detection pipelines that already produce multiple output heads.
  • Edge deployment in autonomous driving or surveillance may become more practical once the accuracy-speed-size frontier moves as described.

Load-bearing premise

The two observed problems—suboptimal pseudo-labels from single-head or naive combination and collapsed multi-scale feature discriminability—are the main reasons vanilla mean-teacher fails on dual-head detectors, and that DHF and MARD fix them without creating new adaptation problems.

What would settle it

An ablation that replaces DHF with either single-head or direct combination pseudo-labeling and removes MARD, then measures whether mAP on a domain-shift benchmark drops back to the level of vanilla mean-teacher.

Figures

Figures reproduced from arXiv: 2606.31834 by Muhammad Haris Khan, Poornima Jain, Sairam VCR, Varun Gopal, Vineeth N Balasubramanian.

Figure 1
Figure 1. Figure 1: Accuracy-speed-size trade-off in SFOD. We compare state-of-the￾art SFOD methods on Cityscapes → Foggy Cityscapes domain-shift, in terms of mAP (%) and inference speed (FPS), with model size (# parameters) encoded by color-graded bubble/star size. Our method achieves the best trade-off across all model-scales, obtaining 53.8% mAP with the large model while remaining the fastest (233 FPS) and most compact (7… view at source ↗
Figure 2
Figure 2. Figure 2: Motivation for our proposed modules on the Cityscapes → Foggy Cityscapes benchmark. (a) O2O pseudo-labels are precise but miss many ob￾jects; O2M pseudo-labels with standard post-processing provide broader cover￾age but introduce additional noise; our proposed DHF achieves the best F1-score. (b) Effective rank of YOLOv10 multi-scale PAN features (P3, P4, P5) under domain-shift. The dashed blue line marks t… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of RT-SFOD. (i) Mean-Teacher framework with DHF and MARD. (ii) DHF: High-precision O2O predictions serve as anchors; non￾redundant O2M boxes with low overlap w.r.t. O2O anchors are selectively added to form the final pseudo-labels. (iii) MARD: (iii-a) Foreground and back￾ground feature vectors are sampled from multi-scale features using pseudo boxes and regularized via variance and covariance cons… view at source ↗
Figure 4
Figure 4. Figure 4: Hyperparameter sensitivity on C2F (RT-SFOD-M). mAP (%) vs. O2O threshold τo2o, O2M threshold τo2m, MARD weight λ0, and EMA momentum µ. Variation is ≤1.5 mAP in all cases. Hyperparameter sensitiv￾ity [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The average cosine similarity of each anchor point’s extracted features with all others across PAN scales(P3, P4, P5) on Foggy Cityscapes validation set. Feature Discriminability Analy￾sis [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
read the original abstract

Real-world detectors for autonomous driving, surveillance, and robotics must handle domain-shifts under strict latency and memory constraints, yet existing source-free object detection (SFOD) methods rely on heavyweight architectures that prioritize accuracy alone. We show this trade-off is unnecessary: building on YOLOv10, an NMS-free dual-head detector, we achieve state-of-the-art adaptation accuracy while being faster and more compact. We observe that directly applying vanilla mean-teacher self-training to dual-head detectors leads to suboptimal adaptation performance due to two key factors. First, simple pseudo-label generation strategies, such as using a single head or directly combining high-confidence predictions from both heads, yield suboptimal supervision under domain-shift. We propose DHF (Dual-Head Pseudo-Label Fusion) which selectively admits one-to-one (O2O) and one-to-many (O2M) head predictions, preserving precision and recovering missed objects. Second, we observe domain-shift collapses multi-scale feature discriminability. We propose the use of our MARD (Multi-scale Adaptive Representation Diversification) loss which mitigates this by enforcing detection-aware variance and covariance constraints on multi-scale feature maps. Both modules are training-time only, leaving inference unchanged. Across domain-shift benchmarks, our method, RT-SFOD yields 1.4 to 3.5\% mAP gains, 1.3$\times$ higher throughput, with $\sim$2$\times$ fewer parameters than prior state-of-the-art SFOD methods, thus advancing the Pareto frontier of the speed-accuracy-model size trade-off. We report main results with YOLOv10, and demonstrate generalizability with additional YOLO- and DETR-based dual-head detectors. Code is available here: https://github.com/Sairam13001/RT-SFOD/

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces RT-SFOD, a source-free object detection method built on the YOLOv10 dual-head detector. It identifies two limitations of vanilla mean-teacher self-training under domain shift—suboptimal pseudo-label generation from single-head or naive fusion strategies, and collapsed multi-scale feature discriminability—and proposes DHF (Dual-Head Pseudo-Label Fusion) to selectively combine O2O and O2M predictions plus the MARD loss to enforce detection-aware variance and covariance constraints on multi-scale features. Both modules are training-only. The central empirical claim is 1.4–3.5% mAP improvement, 1.3× higher throughput, and ~2× fewer parameters versus prior SFOD methods on domain-shift benchmarks, with generalization shown on additional YOLO and DETR dual-head detectors and code released.

Significance. If the reported gains are robust, the work meaningfully advances the speed-accuracy-model-size Pareto frontier for practical SFOD, which is relevant for latency-sensitive applications. The explicit code release supports reproducibility, a clear strength.

major comments (2)
  1. [Abstract, §4] Abstract and §4 (results): The attribution of the 1.4–3.5% mAP gains specifically to DHF and MARD rests on the premise that these modules correct the two identified factors without side effects, yet no controlled ablation is described that replaces the pseudo-label strategy or removes MARD while holding the remainder of the mean-teacher pipeline fixed to show the performance delta disappears.
  2. [§3.1–3.2] §3.1–3.2: The observations that single-head/naive fusion yields suboptimal supervision and that domain shift collapses multi-scale discriminability are presented as primary causes, but the manuscript provides no quantitative isolation (e.g., teacher-student consistency metrics or feature variance statistics before/after each module) demonstrating these are the dominant failure modes rather than other adaptation pathologies.
minor comments (2)
  1. [Tables 1–3, §4] Table captions and experimental setup paragraphs should explicitly list the exact baselines, number of runs, and whether error bars reflect standard deviation across seeds.
  2. [§3.1] Notation for O2O and O2M in the DHF description could be clarified with a short equation or diagram to avoid ambiguity in how predictions are selectively admitted.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments below and will incorporate additional experiments to strengthen the empirical support for our claims.

read point-by-point responses

  1. Referee: [Abstract, §4] Abstract and §4 (results): The attribution of the 1.4–3.5% mAP gains specifically to DHF and MARD rests on the premise that these modules correct the two identified factors without side effects, yet no controlled ablation is described that replaces the pseudo-label strategy or removes MARD while holding the remainder of the mean-teacher pipeline fixed to show the performance delta disappears.

    Authors: We agree that the manuscript would benefit from a more tightly controlled ablation that holds the mean-teacher self-training pipeline fixed while varying only the pseudo-label generation strategy (e.g., single-head vs. naive fusion vs. DHF) or ablating MARD. The current results demonstrate gains relative to prior SFOD methods and include component-level studies, but do not isolate the contributions in exactly the manner described. In the revised manuscript we will add these controlled experiments and report the resulting performance deltas. revision: yes

  2. Referee: [§3.1–3.2] §3.1–3.2: The observations that single-head/naive fusion yields suboptimal supervision and that domain shift collapses multi-scale discriminability are presented as primary causes, but the manuscript provides no quantitative isolation (e.g., teacher-student consistency metrics or feature variance statistics before/after each module) demonstrating these are the dominant failure modes rather than other adaptation pathologies.

    Authors: The observations were derived from our development-stage analysis of dual-head detectors under domain shift. To provide the requested quantitative isolation, we will include additional metrics in the revision, such as teacher-student prediction consistency scores and multi-scale feature variance/covariance statistics computed before and after each module, to more rigorously demonstrate that these are the dominant issues addressed by DHF and MARD. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical claims rest on external benchmarks

full rationale

The paper motivates DHF and MARD from observed issues with vanilla mean-teacher on dual-head detectors, then validates via mAP, throughput, and parameter comparisons on domain-shift benchmarks. No equations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text. All load-bearing claims reduce to experimental results rather than definitions or internal fits, making the work self-contained against external data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No explicit free parameters, axioms, or invented entities described in the abstract beyond standard self-training and loss terms.

pith-pipeline@v0.9.1-grok · 5869 in / 1100 out tokens · 23350 ms · 2026-07-01T05:53:26.675783+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

63 extracted references · 11 canonical work pages · 5 internal anchors

  1. [1]

    Annual Review of Control, Robotics, and Autonomous Systems9(2025)

    Aljalbout, E., Xing, J., Romero, A., Akinola, I., Garrett, C.R., Heiden, E., Gupta, A., Hermans, T., Narang, Y., Fox, D., et al.: The reality gap in robotics: Chal- lenges, solutions, and best practices. Annual Review of Control, Robotics, and Autonomous Systems9(2025)

    2025

  2. [2]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion

    Ashraf, T., Bashir, J.: Titan: Query-token based domain adaptive adversarial learn- ing. In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion. pp. 250–262 (2025)

    2025

  3. [3]

    VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning

    Bardes, A., Ponce, J., LeCun, Y.: Vicreg: Variance-invariance-covariance regular- ization for self-supervised learning. arXiv preprint arXiv:2105.04906 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  4. [4]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Chen, C., Zheng, Z., Ding, X., Huang, Y., Dou, Q.: Harmonizing transferability and discriminability for adapting object detectors. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8869–8878 (2020)

    2020

  5. [5]

    IEEE Transac- tions on Image Processing34, 982–994 (2025)

    Chen, L., Han, J., Wang, Y.: Datr: Unsupervised domain adaptive detection trans- former with dataset-level adaptation and prototypical alignment. IEEE Transac- tions on Image Processing34, 982–994 (2025)

    2025

  6. [6]

    arXiv preprint arXiv:2406.03459 (2024) 4, 10

    Chen, Q., Su, X., Zhang, X., Wang, J., Chen, J., Shen, Y., Han, C., Chen, Z., Xu, W., Li, F., et al.: Lw-detr: A transformer replacement to yolo for real-time detection. arXiv preprint arXiv:2406.03459 (2024)

    work page arXiv 2024

  7. [7]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Chen, Y., Li, W., Sakaridis, C., Dai, D., Van Gool, L.: Domain adaptive faster r-cnn for object detection in the wild. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3339–3348 (2018)

    2018

  8. [8]

    Proceedings of the AAAI Conference on Artificial Intelligence 37(1), 452–460 (Jun 2023)

    Chu, Q., Li, S., Chen, G., Li, K., Li, X.: Adversarial alignment for source free object detection. Proceedings of the AAAI Conference on Artificial Intelligence 37(1), 452–460 (Jun 2023)

    2023

  9. [9]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Cordts, M., Omran, M., Ramos, S., Rehfeld, T., Enzweiler, M., Benenson, R., Franke, U., Roth, S., Schiele, B.: The cityscapes dataset for semantic urban scene understanding. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3213–3223 (2016)

    2016

  10. [10]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Deng, J., Li, W., Chen, Y., Duan, L.: Unbiased mean teacher for cross-domain ob- ject detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 4091–4101 (June 2021)

    2021

  11. [11]

    The international journal of robotics research32(11), 1231–1237 (2013)

    Geiger, A., Lenz, P., Stiller, C., Urtasun, R.: Vision meets robotics: The kitti dataset. The international journal of robotics research32(11), 1231–1237 (2013)

    2013

  12. [12]

    In: European Conference on Computer Vision

    Hao, Y., Forest, F., Fink, O.: Simplifying source-free domain adaptation for object detection: Effective self-training strategies and performance insights. In: European Conference on Computer Vision. pp. 196–213. Springer (2024)

    2024

  13. [13]

    MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

    Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., An- dreetto, M., Adam, H.: Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  14. [14]

    Neurocomputing661, 131957 (2026)

    Huo, Y., Wu, T., Shen, Y., Li, X., Tao, Z., Yang, D.: Qrt-detr: Post-training quan- tization for real-time detection transformer. Neurocomputing661, 131957 (2026)

    2026

  15. [15]

    com/ultralytics/ultralytics

    Jocher, G., Chaurasia, A., Qiu, J.: Ultralytics yolov8 (2023),https://github. com/ultralytics/ultralytics

    2023

  16. [16]

    Johnson-Roberson, M., Barto, C., Mehta, R., Sridhar, S.N., Rosaen, K., Vasudevan, R.: Driving in the matrix: Can virtual worlds replace human-generated annotations for real world tasks? arXiv preprint arXiv:1610.01983 (2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  17. [17]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Kennerley, M., Wang, J.G., Veeravalli, B., Tan, R.T.: Cat: Exploiting inter-class dynamics for domain adaptive object detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 16541–16550 (2024) Real-Time Source-Free Object Detection 17

    2024

  18. [18]

    YOLOv11: An Overview of the Key Architectural Enhancements

    Khanam, R., Hussain, M.: Yolov11: An overview of the key architectural enhance- ments. arXiv preprint arXiv:2410.17725 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  19. [19]

    In: European Conference on Computer Vision

    Khanh, T.L.B., Nguyen, H.H., Pham, L.H., Tran, D.N.N., Jeon, J.W.: Dynamic retraining-updating mean teacher for source-free object detection. In: European Conference on Computer Vision. pp. 328–344. Springer (2024)

    2024

  20. [20]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, H., Zhang, R., Yao, H., Zhang, X., Hao, Y., Song, X., Peng, S., Zhao, Y., Zhao, C., Wu, Y., et al.: Seen-da: Semantic entropy guided domain-aware attention for domain adaptive object detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 25465–25475 (2025)

    2025

  21. [21]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Li, S., Ye, M., Zhu, X., Zhou, L., Xiong, L.: Source-free object detection by learn- ing to overlook domain style. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 8014–8023 (June 2022)

    2022

  22. [22]

    Proceedings of the AAAI Conference on Artificial Intelligence35, 8474–8481 (05 2021)

    Li, X., Chen, W., Xie, D., Yang, S., Yuan, P., Pu, S., Zhuang, Y.: A free lunch for unsupervised domain adaptive object detection without source data. Proceedings of the AAAI Conference on Artificial Intelligence35, 8474–8481 (05 2021)

    2021

  23. [23]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Li, Y.J., Dai, X., Ma, C.Y., Liu, Y.C., Chen, K., Wu, B., He, Z., Kitani, K., Vajda, P.: Cross-domain adaptive teacher for object detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7581–7590 (2022)

    2022

  24. [24]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (2023)

    Liu, Q., Lin, L., Shen, Z., Yang, Z.: Periodically exchange teacher-student for source-free object detection. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (2023)

    2023

  25. [25]

    In: International Conference on Learning Representations (2022)

    Liu, S., Li, F., Zhang, H., Yang, X., Qi, X., Su, H., Zhu, J., Zhang, L.: DAB-DETR: Dynamic anchor boxes are better queries for DETR. In: International Conference on Learning Representations (2022)

    2022

  26. [26]

    In: Proceedings of the European Conference on Computer Vision (ECCV)

    Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S.E., Fu, C.Y., Berg, A.C.: Ssd: Single shot multibox detector. In: Proceedings of the European Conference on Computer Vision (ECCV). Lecture Notes in Computer Science, vol. 9905, pp. 21–37. Springer (2016)

    2016

  27. [27]

    arXiv preprint arXiv:2406.05800 (2024)

    Ma, C., Wang, N., Zhao, Z., Chen, Q.A., Shen, C.: Slowperception: Physical-world latency attack against visual perception in autonomous driving. arXiv preprint arXiv:2406.05800 (2024)

    work page arXiv 2024

  28. [28]

    Journal of Applied Informatics and Computing9, 3810–3820 (12 2025)

    Naufaldihanif, R., Kurniawan, D., Tania, K.: Performance analysis of yolo, faster r-cnn, and detr for automated personal protective equipment detection. Journal of Applied Informatics and Computing9, 3810–3820 (12 2025)

    2025

  29. [29]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Noori, M., Hakim, G.A.V., Osowiechi, D., Shakeri, F., Bahri, A., Yazdanpanah, M., Dastani, S., Ben Ayed, I., Desrosiers, C.: Histopath-c: Towards realistic do- main shifts for histopathology vision-language adaptation. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 4890–4900 (2026)

    2026

  30. [30]

    RangiLyu: Nanodet-plus: Super fast and high accuracy lightweight anchor-free object detection model. (2021)

    2021

  31. [31]

    IEEE transactions on pattern analysis and machine intelligence39(6), 1137–1149 (2016)

    Ren, S., He, K., Girshick, R., Sun, J.: Faster r-cnn: Towards real-time object de- tection with region proposal networks. IEEE transactions on pattern analysis and machine intelligence39(6), 1137–1149 (2016)

    2016

  32. [32]

    Robinson, I., Robicheaux, P., Popov, M., Ramanan, D., Peri, N.: Rf-detr: Neural architecture search for real-time detection transformers (2025)

    2025

  33. [33]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Safdari, R., Nikouei Mahani, M.A., Koohi-Moghadam, M., Bae, K.T.: Mixstyleflow: Domain generalization in medical image segmentation using normalizing flows. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 376–385. Springer (2025) 18 Sairam VCR et al

    2025

  34. [34]

    International Journal of Computer Vision126, 973–992 (2018)

    Sakaridis, C., Dai, D., Van Gool, L.: Semantic foggy scene understanding with synthetic data. International Journal of Computer Vision126, 973–992 (2018)

    2018

  35. [35]

    arXiv preprint arXiv:2402.04466 (2024)

    Sinha, S., Dwivedi, S., Azizian, M.: Towards deterministic end-to-end latency for medical ai systems in nvidia holoscan. arXiv preprint arXiv:2402.04466 (2024)

    work page arXiv 2024

  36. [36]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Sun, T., Segu, M., Postels, J., Wang, Y., Van Gool, L., Schiele, B., Tombari, F., Yu, F.: Shift: a synthetic driving dataset for continuous multi-task domain adaptation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 21371–21382 (2022)

    2022

  37. [37]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition

    Tan, M., Pang, R., Le, Q.V.: Efficientdet: Scalable and efficient object detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition. pp. 10781–10790 (2020)

    2020

  38. [38]

    Advances in neural information processing systems30(2017)

    Tarvainen, A., Valpola, H.: Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. Advances in neural information processing systems30(2017)

    2017

  39. [39]

    urlhttps://github.com/ultralytics/yolov5 (Dec 2020), accessed: [Insert date here]

    Ultralytics: Ultralytics yolov5. urlhttps://github.com/ultralytics/yolov5 (Dec 2020), accessed: [Insert date here]

    2020

  40. [40]

    Ultralytics: YOLO26 (2025),https://github.com/ultralytics/ultralytics

    2025

  41. [41]

    In: European Conference on Computer Vision

    Varailhon, S., Aminbeidokhti, M., Pedersoli, M., Granger, E.: Source-free domain adaptation for yolo object detection. In: European Conference on Computer Vision. pp. 218–235. Springer (2024)

    2024

  42. [42]

    arXiv preprint arXiv:2512.17514 (2025)

    VCR, S., Lalla, R., Dayal, A., Kulkarni, T., Lalla, A., Balasubramanian, V.N., Khan, M.H.: Foundation model priors enhance object focus in feature space for source-free object detection. arXiv preprint arXiv:2512.17514 (2025)

    work page arXiv 2025

  43. [43]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2023)

    Vibashan, V., Oza, P., Patel, V.M.: Instance relation graph guided source-free domain adaptive object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2023)

    2023

  44. [44]

    Advances in neural information processing systems 37, 107984–108011 (2024)

    Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., Han, J., et al.: Yolov10: Real-time end-to-end object detection. Advances in neural information processing systems 37, 107984–108011 (2024)

    2024

  45. [45]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    Wang, C.Y., Bochkovskiy, A., Liao, H.Y.M.: YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    2023

  46. [46]

    In: WACV

    Wang, S., Xia, C., Lv, F., Shi, Y.: Rt-detrv3: Real-time end-to-end object detection with hierarchical dense positive supervision. In: WACV. pp. 1628–1636 (2025)

    2025

  47. [47]

    In: Pro- ceedings of the 29th ACM International Conference on Multimedia

    Wang, W., Cao, Y., Zhang, J., He, F., Zha, Z.J., Wen, Y., Tao, D.: Exploring sequence feature alignment for domain adaptive detection transformers. In: Pro- ceedings of the 29th ACM International Conference on Multimedia. p. 1730–1738. MM ’21, Association for Computing Machinery, New York, NY, USA (2021)

    2021

  48. [48]

    Weng, W., Yuan, C.: Mean teacher detr with masked feature alignment: a robust domain adaptive detection transformer framework. In: Proceedings of the Thirty- Eighth AAAI Conference on Artificial Intelligence and Thirty-Sixth Conference on Innovative Applications of Artificial Intelligence and Fourteenth Symposium on Educational Advances in Artificial Inte...

    2024

  49. [49]

    In: Proceedings of the IEEE/CVF interna- tional conference on computer vision

    Wu, A., Liu, R., Han, Y., Zhu, L., Yang, Y.: Vector-decomposed disentanglement for domain-invariant object detection. In: Proceedings of the IEEE/CVF interna- tional conference on computer vision. pp. 9342–9351 (2021)

    2021

  50. [50]

    Advances in Neural Information Processing Systems35, 4203–4217 (2022) Real-Time Source-Free Object Detection 1

    Yang, J., Li, C., Dai, X., Gao, J.: Focal modulation networks. Advances in Neural Information Processing Systems35, 4203–4217 (2022) Real-Time Source-Free Object Detection 1

    2022

  51. [51]

    arXiv preprint arXiv:2511.07301 (2025)

    Yao, H., Zhao, S., Li, P., Cui, Y., Lu, S., Guo, W., Lu, Y., Xu, Y., Xiong, H.: Beyond boundaries: Leveraging vision foundation models for source-free object detection. arXiv preprint arXiv:2511.07301 (2025)

    work page arXiv 2025

  52. [52]

    IEEE Transactions on Image Processing34, 5948–5963 (2025)

    Yao, H., Zhao, S., Lu, S., Chen, H., Li, Y., Liu, G., Xing, T., Yan, C., Tao, J., Ding, G.: Source-free object detection with detection transformer. IEEE Transactions on Image Processing34, 5948–5963 (2025)

    2025

  53. [53]

    In: European Conference on Computer Vision

    Yoon, I., Kwon, H., Kim, J., Park, J., Jang, H., Sohn, K.: Enhancing source-free domain adaptive object detection with low-confidence pseudo label distillation. In: European Conference on Computer Vision. pp. 337–353. Springer (2024)

    2024

  54. [54]

    arXiv preprint arXiv:1805.046872(5), 6 (2018)

    Yu, F., Xian, W., Chen, Y., Liu, F., Liao, M., Madhavan, V., Darrell, T., et al.: Bdd100k: A diverse driving video database with scalable annotation tooling. arXiv preprint arXiv:1805.046872(5), 6 (2018)

    work page arXiv 2018

  55. [55]

    In: ECCV (2024)

    Yu, J., Liu, J., Wei, X., Zhou, H., Nakata, Y., Gudovskiy, D., Okuno, T., Li, J., Keutzer, K., Zhang, S.: Mttrans: Cross-domain object detection with mean-teacher transformer. In: ECCV (2024)

    2024

  56. [56]

    detr: Instructive multi-route training for detection transformers

    Zhang, C.B., Zhong, Y., Han, K.: Mr. detr: Instructive multi-route training for detection transformers. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 9933–9943 (June 2025)

    2025

  57. [57]

    2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp

    Zhang, J., Huang, J., Luo, Z., Zhang, G., Zhang, X., Lu, S.: Da-detr: Domain adap- tive detection transformer with information fusion. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp. 23787–23798 (2021)

    2023

  58. [58]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Zhao, C., Sun, Y., Wang, W., Chen, Q., Ding, E., Yang, Y., Wang, J.: Ms-detr: Efficient detr training with mixed supervision. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 17027– 17036 (June 2024)

    2024

  59. [59]

    In: 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., Chen, J.: Detrs beat yolos on real-time object detection. In: 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 16965–16974 (2024)

    2024

  60. [60]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Zhao, Z., Wei, S., Chen, Q., Li, D., Yang, Y., Peng, Y., Liu, Y.: Masked retraining teacher-student framework for domain adaptive object detection. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 19039–19049 (2023)

    2023

  61. [61]

    Zhu, X., Su, W., Lu, L., Li, B., Wang, X., Dai, J.: Deformable detr: Deformable transformers for end-to-end object detection. arXiv preprint arXiv:2010.04159 (2020) Supplementary for Real-Time Source-Free Object Detection Sairam VCR1, Varun Gopal1, Poornima Jain 1, Vineeth N Balasubramanian 1,2, and Muhammad Haris Khan 3 1IIT Hyderabad, India 2Microsoft R...

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  62. [62]

    is a large-scale driving dataset with 100,000 images collected under diverse environmental and weather conditions. S.4.3 Details of Data Augmentation Strategies To facilitate robust self-training in our Source-Free Object Detection (SFOD) framework, we employ a Teacher-Student distillation mechanism with asymmet- Real-Time Source-Free Object Detection 7 r...

  63. [63]

    However, AdaBN does not eliminate pseudo-label noise or the non-stationarity introduced by augmentation and thresholding

    shows that Adaptive Batch Normalization (AdaBN) provides a better start- ing point on the target domain and makes fixed pseudo-label training competitive with EMA-based methods. However, AdaBN does not eliminate pseudo-label noise or the non-stationarity introduced by augmentation and thresholding. Figure S.7 studies the effect of teacher update frequency...