pith. sign in

arxiv: 2606.09081 · v1 · pith:LMI43I5Nnew · submitted 2026-06-08 · 💻 cs.CV

Edge-Constrained UAV Small-Object Detection with P2 Enhancement and Quantum-Inspired Lightweight Structure Search

Wuming Lei , Yanbin Gao , Mingyan Sun , Xiaobin Li , Xuechen Liang This is my paper

Pith reviewed 2026-06-27 17:27 UTC · model grok-4.3

classification 💻 cs.CV
keywords UAV object detectionsmall-object detectionedge deploymentP2 branchYOLOX-Nanoquantum-inspired evolutionary algorithmVisDrone
0
0 comments X

The pith

Adding a P2 high-resolution branch to YOLOX-Nano raises small-object AP by 31 percent on VisDrone while meeting edge constraints.

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

The paper tests ways to keep small-object detail in compact UAV detectors that must run on limited onboard hardware. Repeated downsampling in lightweight nets erodes shallow features, so the authors add a P2 detection head that operates at higher resolution. They pair this with a quantum-inspired evolutionary search that screens structures using accuracy, FLOPs, latency, memory, and recall together. On VisDrone the P2 change lifts APsmall 31.10 percent above the YOLOX-Nano baseline and beats a similar-size NanoDet-Plus by 44.9 percent in the same metric. Full 100-epoch checks reveal that the search proxies do not always predict final accuracy rankings.

Core claim

The central finding is that a P2 high-resolution detection branch added to YOLOX-Nano under edge constraints improves APsmall by 31.10 percent over the baseline on VisDrone. The same model also raises APs0.ss by 17.5 percent and APsmall by 44.9 percent relative to NanoDet-Plus at comparable size. QIEA screening yields a candidate with the highest recall, yet after full training the +P2 variant remains strongest on AP-oriented measures. Proxy orderings from the search do not reliably carry over to final AP values.

What carries the argument

The P2 high-resolution detection branch that retains shallow spatial information for small objects, paired with QIEA multi-objective screening over accuracy, FLOPs, latency, memory, and recall.

If this is right

  • P2 serves as the main path for small-object gains in edge UAV detectors without added attention or fusion modules.
  • QIEA supplies a practical tool for joint accuracy-cost screening of lightweight structures.
  • Full training verification remains necessary because proxy rankings do not transfer directly to final AP.
  • The approach supports compact detectors that preserve small-object details under onboard memory and compute limits.

Where Pith is reading between the lines

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

  • The same P2 addition could be tested on other lightweight detectors to check whether the 30-plus percent small-object lift generalizes.
  • Extending the search space to include more task-specific heads might further improve recall-oriented variants.
  • If proxy-to-final mismatch persists, a cheaper early-stopping or partial-training predictor could reduce the cost of structure search.

Load-bearing premise

The multi-objective metrics and QIEA proxy rankings will identify structures whose accuracy after 100 epochs of training matches the proxy ordering.

What would settle it

Train the Random-best, GA-best, and SA/QUBO-best candidates for 100 epochs each and check whether their final AP values preserve the same ranking produced by the QIEA proxies.

Figures

Figures reproduced from arXiv: 2606.09081 by Mingyan Sun, Wuming Lei, Xiaobin Li, Xuechen Liang, Yanbin Gao.

Figure 1
Figure 1. Figure 1: Small-object scale distribution and feature-resolution motivation. The left panel compares [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Experimental workflow and evidence chain. The main path proceeds from data preparation [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mean metrics on the main dataset across three seeds. +P2 provides the most stable [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: High-resolution UAV detection visualization. The four panels compare ground truth, [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Proxy search curve under the fixed candidate budget. The horizontal axis denotes the [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Efficiency and deployment cost metrics, including parameters, FLOPs, GPU latency, [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: AU-AIR engineering case results. Target-domain fine-tuning improves cross-dataset adap [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
read the original abstract

Unmanned aerial vehicle (UAV) object detection requires compact detectors that retain small-object details under onboard computation and memory constraints. Repeated downsampling inlightweight networks weakens shallow spatial information, while manually adding attention orfusion modules may increase cost without stable gains. This study analyzes YOLOX-Nano underedge-deployment constraints by combining a P2 high-resolution detection branch with a quantum-inspired evolutionary algorithm (QIEA) for lightweight structure screening. The search space isdefined by lightweight priority and task specificity, and the evaluation jointly considers accuracy,floating-point operations (FLOPs), latency, memory consumption, and recall. On VisDrone, theP2 branch increases APamall by 31.10% over the YOLOX-Nano baseline. Compared with NanoDet-Plus with similar model size, YOLOX-Nano+-P2 improves APs0.ss by 17.5% and APamal by 44.9%.The QIEA-selected candidate obtains the highest Recallso, but +P2 remains the strongest AP-oriented variant after full training. Full 100-epoch verification of Random-best, GA-best, andSA/QUBO-best candidates further shows that proxy rankings do not necessarily transfer to finalAPse9s. These results support using P2 as the main small-object enhancement path and QIEA as alightweight tool for candidate screening and accuracy-cost analysis. The source code, configurationfiles, diagnostic scripts, and summarized results are available at https://github.com/Ming23233/UAV-QIEA-Edge-Detection

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 paper introduces a P2 high-resolution detection branch to YOLOX-Nano for UAV small-object detection under edge constraints and applies a quantum-inspired evolutionary algorithm (QIEA) to screen lightweight network structures using a multi-objective proxy (accuracy, FLOPs, latency, memory, recall). On VisDrone it reports a 31.10% APsmall gain from the P2 branch over the YOLOX-Nano baseline and further gains versus NanoDet-Plus; the QIEA-selected model achieves highest recall but the paper notes that proxy rankings do not transfer to final 100-epoch AP, with +P2 remaining the strongest AP variant. Source code and scripts are released.

Significance. If the P2 branch gains prove robust, the work supplies a concrete, low-cost architectural change for preserving spatial detail in lightweight UAV detectors. The explicit release of code, configurations, diagnostic scripts, and summarized results is a clear strength that supports reproducibility and follow-on work. The QIEA component is presented as a screening tool rather than a guaranteed optimizer, which limits its claimed impact given the paper's own transferability caveat.

major comments (2)
  1. [Abstract] Abstract: the claim that QIEA provides effective 'lightweight structure screening' is undercut by the explicit statement that 'proxy rankings do not necessarily transfer to final APs' after 100-epoch verification of Random/GA/SA/QUBO candidates; this gap between proxy ordering and final AP is load-bearing for the search contribution and requires either stronger justification or additional experiments showing when the proxy is predictive.
  2. [Abstract] Abstract and evaluation sections: concrete percentage lifts (e.g., +31.10% APsmall, +44.9% APsmall vs. NanoDet-Plus) are reported without error bars, number of runs, dataset split details, or statistical tests, making it impossible to judge whether the observed differences exceed run-to-run variance.
minor comments (2)
  1. [Abstract] Abstract contains multiple typographical errors that impair readability: 'inlightweight', 'APamall', 'APs0.ss', 'APamal', 'APse9s'.
  2. The multi-objective evaluation criteria (accuracy + FLOPs + latency + memory + recall) are listed but the precise weighting or aggregation method used inside QIEA is not stated, complicating reproduction of the proxy ranking step.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract claims and evaluation reporting. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that QIEA provides effective 'lightweight structure screening' is undercut by the explicit statement that 'proxy rankings do not necessarily transfer to final APs' after 100-epoch verification of Random/GA/SA/QUBO candidates; this gap between proxy ordering and final AP is load-bearing for the search contribution and requires either stronger justification or additional experiments showing when the proxy is predictive.

    Authors: We agree the transferability gap is important and already note it explicitly in the manuscript. QIEA is positioned as a lightweight screening tool for multi-objective candidate ranking (accuracy, FLOPs, latency, memory, recall) rather than a guaranteed optimizer; the primary contribution remains the P2 branch, which yields the strongest final AP after full training. We will revise the abstract to emphasize this screening role and clarify that proxy-based selection is followed by 100-epoch verification, with +P2 confirmed as the top AP variant. No new experiments are added, but the existing verification results provide the justification. revision: yes

  2. Referee: [Abstract] Abstract and evaluation sections: concrete percentage lifts (e.g., +31.10% APsmall, +44.9% APsmall vs. NanoDet-Plus) are reported without error bars, number of runs, dataset split details, or statistical tests, making it impossible to judge whether the observed differences exceed run-to-run variance.

    Authors: We acknowledge the lack of error bars and statistical tests. All results use the standard VisDrone train/val/test splits with single runs per configuration under fixed seeds, due to edge-device compute constraints. We will update the abstract and evaluation sections to report the number of runs (one per model), add split details, and note the absence of variance analysis as a limitation. Error bars cannot be added without new multi-run experiments. revision: partial

Circularity Check

0 steps flagged

No circularity; all claims are empirical measurements against external baselines with explicit proxy limitations noted

full rationale

The paper reports measured AP improvements from adding a P2 branch and from QIEA-based candidate screening on VisDrone, with direct comparisons to YOLOX-Nano and NanoDet-Plus. No derivation, equation, or uniqueness claim reduces to a fitted parameter or self-citation by construction; the abstract itself states that proxy rankings do not transfer to final 100-epoch APs, confirming the results are falsifiable empirical outcomes rather than tautological. The search procedure is presented as a screening tool whose limitations are disclosed, not as a load-bearing theorem.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, mathematical axioms or newly postulated entities; the method rests on standard convolutional object-detection components and an off-the-shelf evolutionary search variant.

pith-pipeline@v0.9.1-grok · 5834 in / 1209 out tokens · 24839 ms · 2026-06-27T17:27:11.087063+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

65 extracted references · 12 canonical work pages

  1. [1]

    Vision Meets Drones: A Challenge

    Zhu P, Wen L, Bian X, Ling H, and Hu Q. Vision Meets Drones: A Challenge. 2018. arXiv:1804.07437

    Pith/arXiv arXiv 2018

  2. [3]

    DOTA: A Large-Scale Dataset for Object Detection in Aerial Images

    Xia GS, Bai X, Ding J, et al. DOTA: A Large-Scale Dataset for Object Detection in Aerial Images. In:IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. 2018.doi:10 . 1109 / CVPR . 2018.00418

    arXiv 2018

  3. [4]

    VisDrone- DET2021: The Vision Meets Drone Ob- 17 ject Detection Challenge Results

    Du D, Zhu P, Wen L, et al. VisDrone- DET2021: The Vision Meets Drone Ob- 17 ject Detection Challenge Results. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 2021

    2021

  4. [5]

    Object Detection in UAV Im- ages: A Survey

    Hu X et al. Object Detection in UAV Im- ages: A Survey. Remote Sensing 2024

    2024

  5. [6]

    A Survey on Object Detection in Optical Remote Sensing Im- ages

    Cheng G and Han J. A Survey on Object Detection in Optical Remote Sensing Im- ages. ISPRS Journal of Photogrammetry and Remote Sensing 2016

    2016

  6. [7]

    Deep Learning in Remote Sensing: A Compre- hensive Review and List of Resources

    Zhu XX, Tuia D, Mou L, et al. Deep Learning in Remote Sensing: A Compre- hensive Review and List of Resources. IEEE Geoscience and Remote Sensing Magazine 2017

    2017

  7. [8]

    Feature pyramid networks for object detection,

    Lin TY, Dollar P, Girshick R, He K, Har- iharan B, and Belongie S. Feature Pyra- mid Networks for Object Detection. In: IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition. 2017.doi: 10.1109/CVPR.2017.106

    work page doi:10.1109/cvpr.2017.106 2017

  8. [9]

    Focal Loss for Dense Object De- tection

    Lin TY, Goyal P, Girshick R, He K, and Dollar P. Focal Loss for Dense Object De- tection. In:IEEE International Confer- ence on Computer Vision. 2017.doi:10. 1109/ICCV.2017.324

    2017

  9. [10]

    FCOS: Fully Convolutional One-Stage Object De- tection

    Tian Z, Shen C, Chen H, and He T. FCOS: Fully Convolutional One-Stage Object De- tection. In:IEEE International Confer- ence on Computer Vision. 2019.doi:10. 1109/ICCV.2019.00972

    arXiv 2019

  10. [11]

    Deep Learning for Generic Object Detection: A Survey

    Liu L, Ouyang W, Wang X, et al. Deep Learning for Generic Object Detection: A Survey. International Journal of Com- puter Vision 2020

    2020

  11. [12]

    Faster R-CNN: Towards Real-Time Ob- ject Detection with Region Proposal Net- works

    Ren S, He K, Girshick R, and Sun J. Faster R-CNN: Towards Real-Time Ob- ject Detection with Region Proposal Net- works. IEEE Transactions on Pattern Analysis and Machine Intelligence 2017

    2017

  12. [13]

    Phase-field Approaches to Structural Topology Optimization

    Liu W, Anguelov D, Erhan D, et al. SSD: Single Shot MultiBox Detector. In:Eu- ropean Conference on Computer Vision. 2016.doi:10.1007/978- 3- 319- 46448- 0_2

    work page doi:10.1007/978- 2016

  13. [14]

    Libra R-CNN: Towards Balanced Learning for Object Detection

    Pang J, Chen K, Shi J, Feng H, Ouyang W, and Lin D. Libra R-CNN: Towards Balanced Learning for Object Detection. In:IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019. doi:10.1109/CVPR.2019.00091

    work page doi:10.1109/cvpr.2019.00091 2019

  14. [15]

    Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection

    Zhang S, Chi C, Yao Y, Lei Z, and Li SZ. Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection. In:IEEE/CVF Conference on Computer Vision and Pat- tern Recognition. 2020.doi:10 . 1109 / CVPR42600.2020.00978

    arXiv 2020

  15. [16]

    YOLOX: Exceeding YOLO Series in 2021

    Ge Z, Liu S, Wang F, Li Z, and Sun J. YOLOX: Exceeding YOLO Series in 2021

    2021

  16. [17]

    Byte- Track: Multi-Object Tracking by Associ- ating Every Detection Box

    Zhang Y, Sun P, Jiang Y, et al. Byte- Track: Multi-Object Tracking by Associ- ating Every Detection Box. In:European Conference on Computer Vision. 2022. doi:10.1007/978-3-031-20047-2_1

    work page doi:10.1007/978-3-031-20047-2_1 2022

  17. [18]

    Edge Intelligence: Paving the Last Mile of Artificial Intelligence with Edge Computing

    Zhou Z, Chen X, Li E, Zeng L, Luo K, and Zhang J. Edge Intelligence: Paving the Last Mile of Artificial Intelligence with Edge Computing. Proceedings of the IEEE 2019

    2019

  18. [19]

    Convergence of Edge Computing and Deep Learning: A Com- prehensive Survey

    Chen J and Ran X. Convergence of Edge Computing and Deep Learning: A Com- prehensive Survey. IEEE Access 2019

    2019

  19. [20]

    Model Compression and Acceleration for Deep Neural Networks: The Princi- ples, Progress, and Challenges

    Cheng Y, Wang D, Zhou P, and Zhang T. Model Compression and Acceleration for Deep Neural Networks: The Princi- ples, Progress, and Challenges. IEEE Sig- nal Processing Magazine 2018

    2018

  20. [21]

    Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better

    Menghani G. Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better. ACM Com- puting Surveys 2023. 18

    2023

  21. [22]

    Clustered Object Detection in Aerial Images

    Yang F, Fan H, Chu P, Blasch E, and Ling H. Clustered Object Detection in Aerial Images. In:IEEE International Confer- ence on Computer Vision. 2019.doi:10. 1109/ICCV.2019.00840

    arXiv 2019

  22. [23]

    VistrongerDet: Stronger Visual Information for Object Detection in VisDrone Images

    Wan F et al. VistrongerDet: Stronger Visual Information for Object Detection in VisDrone Images. In:IEEE/CVF In- ternational Conference on Computer Vi- sion Workshops. 2021.doi:10 . 1109 / ICCVW54120.2021.00316

    arXiv 2021

  23. [24]

    Small Object Detection in UAV Images Based on YOLOv8n

    Xu L, Zhao Y, Zhai Y, Huang L, and Ruan C. Small Object Detection in UAV Images Based on YOLOv8n. International Jour- nal of Computational Intelligence Systems 2024

    2024

  24. [25]

    TA-YOLO: A Lightweight Small Object Detection Model Based on Multi-Dimensional Trans-Attention Mod- ule for Remote Sensing Images

    Zhao X et al. TA-YOLO: A Lightweight Small Object Detection Model Based on Multi-Dimensional Trans-Attention Mod- ule for Remote Sensing Images. Complex & Intelligent Systems 2024

    2024

  25. [26]

    UAV-YOLO: Small Object Detection on Unmanned Aerial Vehicle Perspective

    Liu M et al. UAV-YOLO: Small Object Detection on Unmanned Aerial Vehicle Perspective. Sensors 2020

    2020

  26. [27]

    Improved YOLOX-X Based UAV Aerial Photography Object Detection Al- gorithm

    Wang X, He N, Hong C, Wang Q, and Chen M. Improved YOLOX-X Based UAV Aerial Photography Object Detection Al- gorithm. Image and Vision Computing 2023

    2023

  27. [28]

    Small Target-YOLOv5: A Lightweight Small Object Detection Al- gorithm for UAV Aerial Images

    Wang Y et al. Small Target-YOLOv5: A Lightweight Small Object Detection Al- gorithm for UAV Aerial Images. Sensors 2024

    2024

  28. [29]

    SOD-YOLO: A Small Object Detection Method for UAV Images

    Liu X et al. SOD-YOLO: A Small Object Detection Method for UAV Images. Re- mote Sensing 2024

    2024

  29. [30]

    HSP-YOLOv8: A High- Speed and Precise Small Object Detection Method for UAV Aerial Images

    Chen X et al. HSP-YOLOv8: A High- Speed and Precise Small Object Detection Method for UAV Aerial Images. Drones 2024

    2024

  30. [31]

    Enhanced YOLOv8 for Small-Object Detection in Multiscale UAV Imagery: Innovations in Detection Accuracy and Efficiency

    Luo W and Yuan S. Enhanced YOLOv8 for Small-Object Detection in Multiscale UAV Imagery: Innovations in Detection Accuracy and Efficiency. Digital Signal Processing 2025

    2025

  31. [32]

    Dron- eVehicle: A Large-Scale Dataset for UAV- Based Vehicle Detection and Counting

    Sun Y, Cao B, Zhu P, and Hu Q. Dron- eVehicle: A Large-Scale Dataset for UAV- Based Vehicle Detection and Counting. IEEE Transactions on Geoscience and Re- mote Sensing 2022

    2022

  32. [33]

    Recent Ad- vances in Small Object Detection Based on Deep Learning: A Review

    Tong K, Wu Y, and Zhou F. Recent Ad- vances in Small Object Detection Based on Deep Learning: A Review. Image and Vision Computing 2020

    2020

  33. [34]

    Deep High-Resolution Representation Learning for Human Pose Estimation

    Sun K, Xiao B, Liu D, and Wang J. Deep High-Resolution Representation Learning for Human Pose Estimation. In: IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition. 2019.doi: 10.1109/CVPR.2019.00584

    work page doi:10.1109/cvpr.2019.00584 2019

  34. [35]

    Deep High-Resolution Representation Learning for Visual Recognition

    Wang J, Sun K, Cheng T, et al. Deep High-Resolution Representation Learning for Visual Recognition. IEEE Transac- tions on Pattern Analysis and Machine In- telligence 2021

    2021

  35. [36]

    Multi-Scale Feature Fusion for Object Detection in UAV Images

    Li Y et al. Multi-Scale Feature Fusion for Object Detection in UAV Images. IEEE Access 2020

    2020

  36. [37]

    Mo- bileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

    Howard AG, Zhu M, Chen B, et al. Mo- bileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

  37. [38]

    MobileNetV2: In- verted Residuals and Linear Bottlenecks

    Sandler M, Howard A, Zhu M, Zhmogi- nov A, and Chen LC. MobileNetV2: In- verted Residuals and Linear Bottlenecks. In:IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2018. doi:10.1109/CVPR.2018.00474

    work page doi:10.1109/cvpr.2018.00474 2018

  38. [39]

    EfficientNet: Rethink- ing Model Scaling for Convolutional Neu- ral Networks

    Tan M and Le QV. EfficientNet: Rethink- ing Model Scaling for Convolutional Neu- ral Networks. In:International Conference on Machine Learning. 2019. arXiv:1905. 11946. 19

    2019

  39. [40]

    Efficient- Det: Scalable and Efficient Object De- tection

    Tan M, Pang R, and Le QV. Efficient- Det: Scalable and Efficient Object De- tection. In:IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion. 2020.doi:10 . 1109 / CVPR42600 . 2020.01079

    arXiv 2020

  40. [41]

    Deep Com- pression: Compressing Deep Neural Net- works with Pruning, Trained Quantiza- tion and Huffman Coding

    Han S, Mao H, and Dally WJ. Deep Com- pression: Compressing Deep Neural Net- works with Pruning, Trained Quantiza- tion and Huffman Coding. In:Interna- tional Conference on Learning Represen- tations. 2016. arXiv:1510.00149

    Pith/arXiv arXiv 2016

  41. [42]

    Barrett, Ron Brightwell, K

    He Y, Lin J, Liu Z, Wang H, Li LJ, and Han S. AMC: AutoML for Model Com- pression and Acceleration on Mobile De- vices. In:European Conference on Com- puter Vision. 2018.doi:10.1007/978-3- 030-01234-2_48

    work page doi:10.1007/978-3- 2018

  42. [43]

    Neu- ral Architecture Search: A Survey

    Elsken T, Metzen JH, and Hutter F. Neu- ral Architecture Search: A Survey. Journal of Machine Learning Research 2019

    2019

  43. [44]

    Designing Neural Networks Through Neuroevolution

    Stanley KO, Clune J, Lehman J, and Mi- ikkulainen R. Designing Neural Networks Through Neuroevolution. Nature Machine Intelligence 2019

    2019

  44. [45]

    ProxylessNAS: Direct Neural Architecture Search on Tar- get Task and Hardware

    Cai H, Zhu L, and Han S. ProxylessNAS: Direct Neural Architecture Search on Tar- get Task and Hardware. In:International Conference on Learning Representations

  45. [46]

    Once-for-All: Train One Network and Specialize it for Efficient Deployment

    Cai H, Gan C, Wang T, Zhang Z, and Han S. Once-for-All: Train One Network and Specialize it for Efficient Deployment. In:International Conference on Learn- ing Representations. 2020. arXiv:1908 . 09791

    2020

  46. [47]

    Mnas- Net: Platform-Aware Neural Architecture Search for Mobile

    Tan M, Chen B, Pang R, et al. Mnas- Net: Platform-Aware Neural Architecture Search for Mobile. In:IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. 2019.doi:10 . 1109 / CVPR . 2019.00293

    arXiv 2019

  47. [48]

    DARTS: Differentiable Architecture Search

    Liu H, Simonyan K, and Yang Y. DARTS: Differentiable Architecture Search. In:In- ternational Conference on Learning Rep- resentations. 2019. arXiv:1806.09055

    Pith/arXiv arXiv 2019

  48. [49]

    FBNet: Hardware-Aware Efficient ConvNet De- sign via Differentiable Neural Architec- ture Search

    Wu B, Dai X, Zhang P, et al. FBNet: Hardware-Aware Efficient ConvNet De- sign via Differentiable Neural Architec- ture Search. In:IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. 2019.doi:10.1109/CVPR.2019. 01100

    work page doi:10.1109/cvpr.2019 2019

  49. [50]

    NAS- FPN: Learning Scalable Feature Pyra- mid Architecture for Object Detection

    Ghiasi G, Lin TY, and Le QV. NAS- FPN: Learning Scalable Feature Pyra- mid Architecture for Object Detection. In: IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition. 2019.doi: 10.1109/CVPR.2019.00721

    work page doi:10.1109/cvpr.2019.00721 2019

  50. [51]

    Quantum-Inspired Evolutionary Algorithm for a Class of Combinatorial Optimization

    Han KH and Kim JH. Quantum-Inspired Evolutionary Algorithm for a Class of Combinatorial Optimization. IEEE Trans- actions on Evolutionary Computation 2002

    2002

  51. [52]

    Quantum-Inspired Evolutionary Algo- rithms for Feature Subset Selection: A Comprehensive Survey

    Vivek Y, Ravi V, and Krishna PR. Quantum-Inspired Evolutionary Algo- rithms for Feature Subset Selection: A Comprehensive Survey. Preprint. 2024

    2024

  52. [53]

    Feature Selection on Quantum Computers

    Muecke S, Heese R, Mueller S, Wolter M, and Piatkowski N. Feature Selection on Quantum Computers. Quantum Machine Intelligence 2023

    2023

  53. [54]

    An Quantum-Inspired Evolutionary Algo- rithm Applied to Design Optimizations of Electromagnetic Devices

    Zhang W, Xu H, Bai Y, and Yang S. An Quantum-Inspired Evolutionary Algo- rithm Applied to Design Optimizations of Electromagnetic Devices. International Journal of Applied Electromagnetics and Mechanics 2012

    2012

  54. [55]

    Scientometric Anal- ysis of Quantum-Inspired Metaheuristic Algorithms

    Pooja and Sood SK. Scientometric Anal- ysis of Quantum-Inspired Metaheuristic Algorithms. Artificial Intelligence Review 2024. 20

    2024

  55. [56]

    Quantum-Inspired Evolutionary Al- gorithm Applied to Neural Architecture Search

    Szwarcman D, Civitarse D, and Vellasco M. Quantum-Inspired Evolutionary Al- gorithm Applied to Neural Architecture Search. Applied Soft Computing 2022

    2022

  56. [57]

    Quantum Bridge Analytics I: A Tutorial on Formulating and Using QUBO Models

    Glover F, Kochenberger G, and Du Y. Quantum Bridge Analytics I: A Tutorial on Formulating and Using QUBO Models. 4OR 2019

    2019

  57. [58]

    Squeeze-and- Excitation Networks

    Hu J, Shen L, and Sun G. Squeeze-and- Excitation Networks. In:IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. 2018.doi:10 . 1109 / CVPR . 2018.00745

    arXiv 2018

  58. [59]

    Coordi- nate Attention for Efficient Mobile Net- work Design

    Hou Q, Zhou D, and Feng J. Coordi- nate Attention for Efficient Mobile Net- work Design. In:IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. 2021.doi:10 . 1109 / CVPR46437 . 2021.01350

    arXiv 2021

  59. [60]

    CBAM: Convolutional Block Atten- tion Module

    Woo S, Park J, Lee JY, and Kweon IS. CBAM: Convolutional Block Atten- tion Module. In:European Conference on Computer Vision. 2018.doi:10 . 1007 / 978-3-030-01234-2_1

    2018

  60. [61]

    Deformable Convolutional Networks

    Dai J, Qi H, Xiong Y, et al. Deformable Convolutional Networks. In:IEEE Inter- national Conference on Computer Vision. 2017.doi:10.1109/ICCV.2017.89

    work page doi:10.1109/iccv.2017.89 2017

  61. [62]

    Detec- toRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution

    Qiao S, Chen LC, and Yuille A. Detec- toRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution. In:IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. 2021.doi:10 . 1109 / CVPR46437 . 2021.01008

    arXiv 2021

  62. [63]

    In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Liu S, Qi L, Qin H, Shi J, and Jia J. Path Aggregation Network for Instance Segmentation. In:IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. 2018.doi:10.1109/CVPR.2018. 00913

    work page doi:10.1109/cvpr.2018 2018

  63. [64]

    End-to- End Object Detection with Transformers

    Carion N, Massa F, Synnaeve G, Usunier N, Kirillov A, and Zagoruyko S. End-to- End Object Detection with Transformers. In:European Conference on Computer Vi- sion. 2020.doi:10 . 1007 / 978 - 3 - 030 - 58452-8_13

    2020

  64. [65]

    Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P

    Liu Z, Lin Y, Cao Y, et al. Swin Trans- former: Hierarchical Vision Transformer Using Shifted Windows. In:IEEE/CVF International Conference on Computer Vision. 2021.doi:10.1109/ICCV48922. 2021.00986

    work page doi:10.1109/iccv48922 2021

  65. [66]

    AU-AIR: A Multi-modal Unmanned Aerial Ve- hicle Dataset for Low Altitude Traffic Surveillance

    Bozcan I and Kayacan E. AU-AIR: A Multi-modal Unmanned Aerial Ve- hicle Dataset for Low Altitude Traffic Surveillance. In:IEEE/CVF Conference on Computer Vision and Pattern Recog- nition Workshops. 2020.doi:10 . 1109 / CVPRW50498.2020.00175. 21

    arXiv 2020