The DOI matches its Crossref record; 26 of the 26 references it cites were found in a registry or an index; 4 of the 23 sources its claims cite were checked for support.
Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection
Open accessAbstract
Deploying large object detectors on embedded edge platforms is governed by a joint trade-off among detection accuracy, end-to-end throughput, and total system power. This paper benchmarks the large variants YOLOv8l and RT-DETR-l across heterogeneous deployment runtimes on two edge platforms: Raspberry Pi 5 with CPU and NPU offload, and Nvidia Jetson Orin NX with GPU acceleration. Accuracy is evaluated on COCO val2017 using mAP50-95, while throughput and energy efficiency (FPS/W) are measured on a realistic end-to-end video pipeline that includes decoding, preprocessing, inference, and post-processing. Model-execution latency is analyzed separately from pipeline throughput to avoid ambiguity between inference time and end-to-end processing rate. On Raspberry Pi 5, CPU-only execution of large models is impractical due to multi-second per-frame latency, whereas NPU acceleration substantially improves energy efficiency for YOLOv8l, albeit with deployment constraints that can reduce accuracy. On Jetson Orin NX, TensorRT provides the strongest deployment path for both architectures; however, the relative ranking of YOLOv8l and RT-DETR-l depends on runtime realization rather than on nominal FLOPs alone. A generalized interpretation based on normalized latency and energy per nominal GFLOP, together with a decomposition of conversion sensitivity and quantization sensitivity, shows that deployment efficiency is jointly determined by nominal compute demand, memory-system behavior, runtime overhead, and accuracy retention after export and quantization. Under the tested conditions, none of the evaluated large-model configurations reaches the strict real-time target of 25 FPS for the full pipeline, indicating that further hardware-specific optimization and/or smaller model variants are still required for real-time edge deployment.
Overview
CoChat Summary
TL;DR
This paper benchmarks large YOLOv8l and RT-DETR-l across edge runtimes on Raspberry Pi 5 and Nvidia Jetson Orin NX, measuring COCO accuracy and end-to-end FPS/W on a realistic video pipeline. It finds deployment efficiency depends jointly on architecture, runtime, hardware, and post-training conversion/quantization, and no evaluated large-model configuration reached the strict ≥25 FPS real-time target.
Key findings
- On Raspberry Pi 5, CPU-only execution of large detectors is impractical due to multi-second per-frame latency, with efficiency dropping as low as 0.0313 FPS/W. Hailo-8 gives the highest measured efficiency for YOLOv8l at 1.7281 FPS/W, and HailoRT preserves 0.541 mAP50–95; Edge TPU exceeds 1.3 FPS/W but falls to 0.324 mAP50–95 because of mandatory INT8 quantization and reduced 224 × 224 input resolution.
- On Jetson Orin NX, TensorRT with INT8 quantization achieves the highest efficiency: 0.9960 FPS/W for YOLOv8l and 1.0597 FPS/W for RT-DETR-l, with sub-100 ms inference times for the evaluated large-model configurations and approximately 90% GPU utilization for both models.
- The ranking is runtime-dependent: YOLOv8l is faster under PyTorch, whereas under TensorRT RT-DETR-l achieves the highest throughput among evaluated Jetson configurations. RT-DETR-l benefits more from compiled execution, but remains less efficient per nominal GFLOP than YOLOv8l; YOLOv8l maintains lower latency and lower energy per nominal workload together with stronger cross-runtime accuracy stability.
- Accuracy loss decomposition shows RT-DETR-l has a larger relative loss in the PyTorch-to-TensorRT FP32 step, while YOLOv8l loses a larger fraction in the subsequent FP32-to-INT8 step; the larger overall PyTorch-to-TensorRT INT8 reduction for RT-DETR-l reflects combined conversion and quantization effects. None of the evaluated configurations reached ≥25 FPS.
Methods
- Benchmarked pre-trained YOLOv8l and RT-DETR-l. Accuracy was evaluated on COCO val2017 (5,000 images, 80 categories) using mAP50–95, mAP50, and mAP75. Throughput and energy efficiency were measured on the MOT16-13 benchmark video (750 frames, 1920 × 1080 at 25 FPS) using an end-to-end pipeline including video decoding, preprocessing, inference, and post-processing.
- Platforms: Raspberry Pi 5 (8GB) running Debian GNU/Linux 12 (kernel 6.6.62) with Coral USB Accelerator and Raspberry Pi AI HAT+; Nvidia Jetson Orin NX (16GB) running L4T 36.4.0 with JetPack 6.1 in MAXN power mode. Software included Ultralytics 8.3, TensorFlow 2.17/2.16, PyTorch 2.5, and specialized runtimes (TensorRT, EdgeTPU, HailoRT, NCNN). Standard factory configurations were used without overclocking or jetson_clocks.
- Power was measured with an OWON SPE8205 programmable DC power supply and a Fnirsi FNB48P USB power meter, reporting Total System Power. Energy efficiency is FPS/W = FPS / average power. Model-execution latency was analyzed separately from end-to-end throughput; real-time was defined as sustained end-to-end throughput ≥25 FPS, corresponding to ≤40 ms per frame.
Limitations
- Stated: the generalized interpretation is an analytical model for interpreting measured results, not a direct profiler-based attribution of every kernel-level contribution. No configuration reached ≥25 FPS, so further hardware-specific optimization, reduced pipeline overhead, and/or smaller detector variants remain necessary. RT-DETR could not be deployed on Edge TPU or Hailo-8 due to conversion and operator-support limitations for transformer-based architectures.
- Evident from the text: the study evaluates only two large models (YOLOv8l and RT-DETR-l) on two edge platforms, uses one benchmark video (MOT16-13) for throughput/energy, and does not fine-tune models, instead exporting pre-trained weights directly. Configurations with mAP50–95 below 0.50 are treated as accuracy-limited for practical use, and standard factory configurations may not reflect fully optimized deployments.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures










Verification
Claims the paper makes
4 Valid · 2 Not valid · 11 Not checked · 57 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“The model’s hybrid encoder integrates Intra-scale Feature Interaction (AIFI) with the CNN-based Cross-scale Feature Fusion (CCFF), resulting in a significant improvement in detection accuracy9.”
§ Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection > Methodology > Pre-trained models YOLOv8 and RT-DETR
What the sources say
- Supportedcites [9] Zhao, Y. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2024, pp. 16965–16974, (2024). https://doi.org/10.1109/CVPR52733.2024.01605 source
Section 4.2 (p. 4) identifies AIFI and CNN-based CCFF as the two modules of RT-DETR’s hybrid encoder. The hybrid-encoder ablation in Section 5.2 (p. 7) reports that the final variant improves detection accuracy by 1.5 AP over its predecessor, supporting the claimed accuracy improvement. (LLM, conf=0.91)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In standard dense self-attention, the attention-score tensor scales as”
§ Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection > Generalized interpretation of architectural efficiency > Architecture‑level sources of latency and energy
What the sources say
- Supportedcites [30] Shehzadi, T., Hashmi, K. A., Liwicki, M., Stricker, D. & Afzal, M. Z. Object Detection with Transformers: A Review. Sensors 25(19), 6025. https://doi.org/10.3390/s25196025 (2025). source
Section 3.2 (Deformable-DETR) gives standard encoder self-attention complexity as O(wᵢ²hᵢ²cᵢ) and defines attention weights over query–key pairs. With N = wᵢhᵢ spatial tokens, the number of attention scores scales as O(N²). The paper uses N separately for decoder object queries; the claim uses it generically for token count. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“RT-DETR-l, in contrast, combines convolutional feature extraction with transformer-based interaction through the hybrid encoder based on AIFI and CCFF9.”
§ Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection > Generalized interpretation of architectural efficiency > Architecture‑level sources of latency and energy
What the sources say
- Supportedcites [9] Zhao, Y. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2024, pp. 16965–16974, (2024). https://doi.org/10.1109/CVPR52733.2024.01605 source
The Model Overview (p. 4) describes RT-DETR as having a backbone and an efficient hybrid encoder. Section 4.2 (pp. 4–5) specifies that the encoder combines Transformer-based AIFI with CNN-based CCFF, supporting the description of convolutional feature extraction and transformer-based interaction. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“RT-DETR could not be deployed on Edge TPU or Hailo-8 due to conversion and operator-support limitations for transformer-based architectures (see Section III-D).”
- Also stated in: § Experimental pipeline · hedged
“RT-DETR could not be deployed on Edge TPU or Hailo-8 due to conversion and operator-support limitations for transformer-based architectures”
§ Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection > Methodology > Experimental pipeline
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Experimental pipeline · hedged
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On Raspberry Pi 5, shown in Figs. 6 and 7, the Hailo-8 accelerator delivers the highest measured efficiency for YOLOv8l, reaching 1.7281 FPS/W and establishing the strongest practical operating point on that platform under the tested conditions.”
- Also stated in: § Energy efficiency · plain
“the Hailo-8 accelerator delivers the highest measured efficiency for YOLOv8l, reaching 1.7281 FPS/W and establishing the strongest practical operating point on that platform under the tested conditions.”
- Also stated in: § Conclusion · plain
“NPU-assisted paths improve FPS/W, with HailoRT as the strongest for YOLOv8l (Fig. 10).”
§ Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection > Results > Energy efficiency
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Energy efficiency · plain
35 more claims
Create a free account to see all 74 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 14
- Borderline: one of the tests for a substantial claim is unclear. 3
- It may state two separate claims as one. 23
Citations
Cites
DETRs Beat YOLOs on Real-time Object Detection
Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., & Chen, J. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.01605 · Cited by 5,020
YOLOv8: A Novel Object Detection Algorithm with Enhanced Performance and Robustness
Varghese, R., & M., S. (2024) · 2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS) · doi:10.1109/adics58448.2024.10533619 · Cited by 2,007
Deep Learning With Edge Computing: A Review
Chen, J., & Ran, X. (2019) · Proceedings of the IEEE · doi:10.1109/jproc.2019.2921977 · Cited by 1,540
Deep learning for unmanned aerial vehicle-based object detection and tracking: a survey
Wu, X., Li, W., Hong, D., Tao, R., & Du, Q. (2022) · IEEE Geoscience and Remote Sensing Magazine · doi:10.1109/mgrs.2021.3115137 · Cited by 421
A Review of SLAM Techniques and Security in Autonomous Driving
Singandhupe, A., & La, H. M. (2019) · 2019 Third IEEE International Conference on Robotic Computing (IRC) · doi:10.1109/irc.2019.00122 · Cited by 147
An Evaluation of Edge TPU Accelerators for Convolutional Neural Networks
Seshadri, K., Akin, B., Laudon, J., Narayanaswami, R., & Yazdanbakhsh, A. (2022) · 2022 IEEE International Symposium on Workload Characterization (IISWC) · doi:10.1109/iiswc55918.2022.00017 · Cited by 138
Transformer for object detection: Review and benchmark
Li, Y., Miao, N., Ma, L., Shuang, F., & Huang, X. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2023.107021 · Cited by 135
Object Detection with Transformers: A Review
Shehzadi, T., Hashmi, K. A., Liwicki, M., Stricker, D., & Afzal, M. Z. (2025) · Sensors · doi:10.3390/s25196025 · Cited by 93
Benchmarking Deep Learning Models for Object Detection on Edge Computing Devices
Alqahtani, D. K., Cheema, M. A., & Toosi, A. N. (2025) · Lecture Notes in Computer Science · doi:10.1007/978-981-96-0805-8_11 · Cited by 85
Embracing drones and the Internet of drones systems in manufacturing – An exploration of obstacles
Askerbekov, D., Garza‐Reyes, J. A., Ghatak, R. R., Joshi, R., Kandasamy, J., & Nascimento, D. L. D. M. (2024) · Technology in Society · doi:10.1016/j.techsoc.2024.102648 · Cited by 75
Research works printed but not linked: 1
- Vaswani, A. et al. Attention is all you need, in Proceedings of the 31st International Conference on Neural Information Processing Systems, Long Beach, California, USA, pp. 6000–6010. (2017).
Cited by
A Circular Bounding Box Detection Algorithm Based on Improved YOLOv8 for Dense Fruit Images
Zhu, J., & Zheng, X. (2026) · International Conference on Big Data and Privacy Computing (BDPC) · doi:10.1109/bdpc69980.2026.11608083
Rezwan, S., Khan, U. A., Martin Reisslein, B., Cabrera, J. A., Hofmann, P., & Fitzek, F. H. P. (2026) · IEEE Access · doi:10.1109/access.2026.3734116
YOLOv8-EVdet: An emergency vehicle detection method for autonomous vehicles
Loukas, P., Bassir, D., & Amanatiadis, A. A. (2026) · International Journal of Advanced Robotic Systems · doi:10.1177/17298806261486243
Metadata
- Authors
- Suchý, I., & Turčaník, M.
- Affiliations
- Armed Forces Academy (SK)
- Corresponding author
- Turčaník, Michal
- Venue
- Scientific Reports
- Published
- 2026-03-27
- Volume
- 16
- Issue
- 1
- ISSN
- 2045-2322 (ISSN-L)
- PMC
- PMC13039985
- OpenAlex
- W7141095983
- Cited by
- 14
- References
- 26
- Citations by year
- 2026: 14
- FWCI
- 38.50
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0 (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0 (vor, 4 d)
- Locations
- Scientific Reports (publishedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Ministerstvo obrany Slovenskej republiky
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Keywords
- YOLOv8, RT-DETR, energy efficiency, edge deployment, real-time object detection, Raspberry Pi 5, GPU acceleration, TensorRT, model quantization, inference latency
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection. https://app.cochat.ai/papers/10.1038/s41598-026-46453-6
- The paper
- Suchý, I., & Turčaník, M. (2026). Review of large YOLOv8 and RT-DETR energy efficiency on edge devices for real-time detection. Scientific Reports. https://doi.org/10.1038/s41598-026-46453-6
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.