Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review
Open accessThe DOI matches its Crossref record; 408 of the 491 references it cites were found in a registry or an index.
Abstract
Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly accurate models for data-driven, learned, automatic, and practical machine learning (ML) solutions to end-user applications remains challenging. DL algorithms are often computationally expensive, power-hungry, and require large memory to process complex and iterative operations of millions of parameters. Hence, training and inference of DL models are typically performed on high-performance computing (HPC) clusters in the cloud. Data transmission to the cloud results in high latency, round-trip delay, security and privacy concerns, and the inability of real-time decisions. Thus, processing on edge devices can significantly reduce cloud transmission cost. Edge devices are end devices closest to the user, such as mobile phones, cyber–physical systems (CPSs), wearables, the Internet of Things (IoT), embedded and autonomous systems, and intelligent sensors. These devices have limited memory, computing resources, and power-handling capability. Therefore, optimization techniques at both the hardware and software levels have been developed to handle the DL deployment efficiently on the edge. Understanding the existing research, challenges, and opportunities is fundamental to leveraging the next generation of edge devices with artificial intelligence (AI) capability. Mainly, four research directions have been pursued for efficient DL inference on edge devices: 1) novel DL architecture and algorithm design; 2) optimization of existing DL methods; 3) development of algorithm–hardware codesign; and 4) efficient accelerator design for DL deployment. This article focuses on surveying each of the four research directions, providing a comprehensive review of the state-of-the-art tools and techniques for efficient edge inference.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“These data cannot be managed by relying only on the limited functional services provided by cloud service centers (Ge et al. 2023;Shuvo et al. 2022).”
From Survey: federated learning data security and privacy-preserving in edge-Internet of Things (2024) · cites this paper as Shuvo et al. 2022
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 1×
“NLP models, such as Transformer [45] and BERT [224], have around 80% [225] and 93% [226] weight sparsity.”
MethodGradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
You Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 1×
“Thus, splitting the computation of convolution and dense layer between GPU and CPU can achieve a 60× speedup for object detection using YOLO in the Jetson TK1 board [283].”
MethodDensely Connected Convolutional Networks
Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.243 · Cited by 48,566
Squeeze-and-Excitation Networks
Hu, J., Shen, L., & Sun, G. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00745 · Cited by 35,667
(2014) · Cited by 33,419
Research works printed but not linked: 30
- Wendelken, 2022, Worldwide and U.S. IoT Cellular Connections Forecast, 2021–2025
- Gupta, ProtoNN: Compressed and accurate KNN for resource-scarce devices, 1331
- Dennis, 2020, EdgeML: Machine learning for resource-constrained edge devices
- Kumar, Resource-efficient machine learning in 2 KB RAM for the Internet of Things, 70, 1935
- 2022, AutoML
- 2022, AutoKeras
- Ryu, BitBlade: Area and energy-efficient precision-scalable neural network accelerator with bitwise summation, 1
- Hojabr, SkippyNN: An embedded stochastic-computing accelerator for convolutional neural networks, 1
- 2022, Jetson TX2 Module
- 2022, Intel® Edison Development Platform
- 2022, Open Neural Network Exchange
- 2022, Deep Learning HDL Toolbox
- 2022, X-CUBE-AI, AI Expansion Pack for STM32CubeMX
- 2022, DNNDK User Guide
- 2022, NVIDIA TensorRT
- 2022, CEVA Deep Neural Network (CDNN)
- 2021, Qualcomm Neural Processing SDK for AI
- 2022, Stratus High-Level Synthesis
- 2022, Deploy Machine Learning Models on Mobile and IoT Devices
- Rasley, DeepSpeed, 3505
- Baines, 2021, FairScale: A general purpose modular PyTorch library for high performance and large scale training
- Dai, BigDL, 50
- 2022, Tensilica DNA Processor Family for on-Device AI
- 2022, DesignWare ARC EV Processors for Embedded Vision
- 2022, AI for the Edge
- 2022, Products, Helping You Bring Local AI to Applications From Prototype to Production
- Shacham, 2022, Pixel Visual Core: Image Processing and Machine Learning on Pixel 2
- Yang, Sparse ReRAM engine: Joint exploration of activation and weight sparsity in compressed neural networks, 236
- Mao, A privacy-preserving deep learning approach for face recognition with edge computing, 1
- Wang, Not just privacy, 2407
Cited by
Items 201–300 of 420
Chen, Q., Wang, H., Zhang, H., Zhu, Z., & Wei, X. (2025) · Life Sciences · doi:10.1016/j.lfs.2025.123931 · Cited by 3
Methodology for optimizing convolutional neural networks for fast production processes
Wree, C., & Raßmann, R. (2025) · at - Automatisierungstechnik · doi:10.1515/auto-2024-0155 · Cited by 2
“The inference engine can be implemented in a cloud (option 1) with highest computational resources [9] or on an edge server (option 2) with high computational resources [2], [10], [11].”
Wang, Q., Tamukoh, H., & Morie, T. (2025) · Frontiers in Neuroscience · doi:10.3389/fnins.2025.1656892 · Cited by 3
Efficient Flexible Edge Inference for Mixed-Precision Quantized DNN using Customized RISC-V Core
Lu, T.-C., Ding, J., Zhang, H., Jiang, B., Xu, W., & Chai, Z. (2025) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3768630 · Cited by 2
“TinyML [12, 30], as a new application direction of DNN deployed on edge computing platforms, aims to enable deep learning workflows to operate effectively under low power and limited computational resources.”
Terahertz downlink channel estimation using AIoT systems: OMP integrated pruned deep CNN method
Ruby, R., Aslam, U., Zhang, D., Wu, K., & Wang, L. (2025) · Computer Networks · doi:10.1016/j.comnet.2025.111654 · Cited by 1
Real Time Anomaly Detection in SDN Using Deep Learning on Edge Enabled Architectures
Bajri, S. A. N., Negara, R. M., & Hertiana, S. N. (2025) · International Conference on Creative Communication and Innovative Technology (ICCIT) · doi:10.1109/iccit65724.2025.11167362 · Cited by 3
“Additionally, little attention has been paid to model optimization techniques such as quantization and compression, which are essential for deploying deep learning models on resource-constrained edge devices in real-world scenarios [10], [11].”
Yekanlou, A., & Cai, J. (2025) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2025.3617852 · Cited by 1
“This area has been well-studied, with works such as [29] presenting various techniques to enhance CR in edge servers for DNN inference.”
Syed, A. A., Sambamurthy, P. K., Mohammed, A. S., et al. (2025) · doi:10.1109/ciacon65473.2025.11189423 · Cited by 1
“The growth estimate to 22 billion devices in 2025 is the size of scalable low-latency solutions [6].”
ON-NSW: Accelerating High-Dimensional Vector Search on Edge Devices with GPU-Optimized NSW
Park, T., Lee, H., Na, Y., & Kim, W.-H. (2025) · Sensors · doi:10.3390/s25206461 · Cited by 2
“Furthermore, edge devices exhibit unique architectural characteristics, such as a limited number of CPU cores and weaker computing power compared to servers [19].”
Wang, S., Wang, Y., Zhang, H., Li, S., Petrany, A., & Chen, Y. (2025) · IEEE Access · doi:10.1109/access.2025.3626650 · Cited by 1
Deep Learning for Real-Time Anomaly Detection in IoT
Soni, A., Subramanya Sarma, S., Shaikh, N., Murugan, S., Lande, J., & Palanisamy, P. (2025) · International Conference on Electronics and Sustainable Communication Systems (ICESC) · doi:10.1109/icesc65114.2025.11212531 · Cited by 1
HydroFedNet: An Intent-Based Unified Federated Framework for Multisource Water Quality Monitoring
Sangaiah, A. K., Mohan, A., Anandakrishnan, J., et al. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3627665 · Cited by 1
“Papers such as [17] showed how RFs predict turbidity and total suspended solids.”
Towards Decentralized IoT Security: Federated Learning with RNNs for Real-Time Anomaly Detection
Arjdal, R., Younes, A., Asimi, A., & Oumouss, L. (2025) · IEEE Congress on Information Science and Technology (CiSt) · doi:10.1109/cist65886.2025.11224260 · Cited by 1
“Furthermore, they rarely account for the resource limitations of edge devices or the presence of non-IID data across the network [6].”
CODA: A Computation-Driven Paradigm for Sparse DNN Acceleration
Liu, Y., Li, W., Zhang, K., et al. (2025) · IEEE Computer Architecture Letters · doi:10.1109/lca.2025.3637756 · Cited by 1
“F OR decades, the spatial dataflow paradigm has been the corner-stone of high-performance computing architecture [1].”
Accelerating GNN Inference via Automated Parallel Execution on Edge Heterogeneous Platforms
Lin, Y.-C., Fan, H., Gobriel, S., Jain, N. K., & Prasanna, V. K. (2025) · International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD) · doi:10.1109/sbac-pad66369.2025.00024 · Cited by 1
“Edge Inference Acceleration: Edge inference [13], [54] is an emerging research area as more applications are empowered by ML and demand low latency.”
Traffic Sign Classification Using Real-Time GPU-Embedded Systems
Lopez-Montiel, M. A., Orozco-Rosas, U., Sánchez-Adame, M., Montiel, O., Picos, K., & Tapia, J. J. (2025) · SN Computer Science · doi:10.1007/s42979-025-04634-6 · Cited by 5
“Therefore, the usage of GPU for edge inference depends on the power and energy handling capacity of the target [18 ].”
InfluentialVemuri, S. M., Gundrapally, A., Xia, Z., Kim, J., & Choi, K. K. (2026) · IEEE Access · doi:10.1109/access.2025.3650071 · Cited by 5
“These systems operate under tight latency, power, and thermal budgets, requiring models that are not only compact but also capable of real-time inference at low voltages and modest clock frequencies [6], [10], [11].”
Li, L. (2026) · Sensor Review · doi:10.1108/sr-07-2025-0499 · Cited by 2
Askr, H., Darwish, A., & Hassanien, A. E. (2026) · Studies in systems, decision and control · doi:10.1007/978-3-032-05701-3_2 · Cited by 2
Distributed quantum model learning for traffic density estimation
Wang, K., Wang, B., Zhai, W., & Cheng, J.-A. (2026) · Image and Vision Computing · doi:10.1016/j.imavis.2026.105900 · Cited by 2
Self-citationA Lightweight Model for Power Quality Disturbance Recognition Targeting Edge Deployment
Bai, H., Yao, R., Liu, T., Ma, Z., Liu, S., Lei, Y., & Zheng, Y. (2026) · Energies · doi:10.3390/en19020368 · Cited by 2
Bio-Inspired Swarm Intelligence for Optimized Energy-Efficient Computing in Smart IoT Ecosystems
Fakhreldin, M. A. I., Jeribi, F., Adam, Y. A., Khawaji, A., Elhadi, H., & Hamdan, A. A. (2025) · International Conference on Software Engineering and Computer Systems · doi:10.1109/icsecs65227.2025.11279170 · Cited by 1
“DL-based approaches offer predictive accuracy but impose high computational complexity, memory requirements, and latency, making them unsuitable for real-time IoT systems [10].”
Liu, W., & Luo, Q. (2026) · IEEE Access · doi:10.1109/access.2026.3654955 · Cited by 3
Yashu, K., Gogi, V. J., & Azarudheen, S. (2025) · IEEE DELCON - International Conference on Recent Smart Technologies in Engineering for Sustainable Development (DELCON) · doi:10.1109/delcon68055.2025.11400330 · Cited by 2
Miyasaka, Y., Hirose, A., & Nakane, R. (2026) · APL Machine Learning · doi:10.1063/5.0308497 · Cited by 1
FPGA-Enabled Deep Learning for Real-Time DC Series Arc Fault Detection
Li, Y., Mao, Y., Rothmann, M., Weiss, R. J., & Porrmann, M. (2026) · IEEE Open Journal of Industry Applications · doi:10.1109/ojia.2026.3671845 · Cited by 3
“Hardware acceleration of DL inference on edge devices is gaining more and more attention [45], [46].”
BitRed: Taming Non-Uniform Bit-Level Sparsity with a Programmable RISC-V ISA for DNN Acceleration
Liu, Y.-H., Li, W.-M., Zhang, K., et al. (2026) · International Conference on Architectural Support for Programming Languages and Operating Systems · doi:10.1145/3779212.3790132 · Cited by 1
“The explosive growth of Deep Neural Networks (DNNs) [24] and Large Language Models (LLMs) [4, 43, 45, 51] imposes formidable computational pressure on hardware accelerators [37, 40].”
Zhang, D., Shi, W. K., & Jia, X. (2026) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2026.3672428 · Cited by 4
UAV-Assisted Edge Inference With Integrated Sensing, Communication, and Computation
Wen, D., Zhang, S., Zhu, G., Liu, Y., Shi, Y., & Hu, H. (2026) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2026.3669999 · Cited by 4
“Specifically, ISCC enables edge learning to distill AI models from vast distributed sensory datasets [4], [5], [6], [7], [8], [9] and supports edge inference to deploy trained AI models for real-time intelligent decisions [10], [11], [12], [13].”
FPGA-Based Front-End Low-Light Enhancement for Deterministic Vision-Only Driving Perception
Xie, F., Jing, H., Lu, Z., Ju, S., Peng, B., Xie, T., Yang, L., Han, W., Wang, Z., & Sai, G. (2026) · Electronics · doi:10.3390/electronics15061224 · Cited by 2
Model Compression for IoT Edge Devices Based on Knowledge Distillation
Lin, K., Lin, Z., & Wang, S. (2026) · International Journal of Pattern Recognition and Artificial Intelligence · doi:10.1142/s0218001426400070 · Cited by 3
DDoS attack detection and classification for the MQTT-IoT protocol using LSTM models
Negesse, D. B., Gemeda, K. A., & Gianini, G. (2026) · Discover Applied Sciences · doi:10.1007/s42452-026-08563-8 · Cited by 3
Mahmoud, H. A., Soliman, A. T., El-Meligy, M. A., et al. (2026) · Sustainable Computing Informatics and Systems · doi:10.1016/j.suscom.2026.101342 · Cited by 1
Analyzing time and power efficiency of machine learning models on edge devices
Khdr, H., Ahmed, M. A., Oğuz, Y., & Henkel, J. (2026) · Analog Integrated Circuits and Signal Processing · doi:10.1007/s10470-026-02580-8 · Cited by 1
Topology modeling and energy efficiency prediction of parallel chillers based on deep learning
Liu, Y., Yang, X., Zhang, L., et al. (2026) · Scientific Reports · doi:10.1038/s41598-026-47180-8 · Cited by 1
“It solves the problem of classifying nodes (such as documents) in graphs (such as citation networks), where only a small number of nodes have labels 33 .”
Alamu, O., Olwal, T. O., & Migabo, E. (2026) · Franklin Open · doi:10.1016/j.fraope.2026.100637 · Cited by 1
DistillSkel: Learning pruned skeletons of 2D shapes through progressive knowledge distillation
Bincy Antony, M., Mishra, A., Raghuram, P., Tamilselvan, D., & Muthuganapathy, R. (2026) · Computers & Graphics · doi:10.1016/j.cag.2026.104647 · Cited by 2
Vision Language Models on the Edge for Real-Time Robotic Perception
Ahmad, S., Hafeez, M., & Zaidi, S. A. R. (2026) · International Conference on Communications · doi:10.1109/icc59461.2026.11587760 · Cited by 9
“Cloud-based inference, which transmits large volumes of sensitive multimodal data, is limited by scalability challenges, excessive wide-area latency, and privacy risks [5].”
Yu, G., Bao, Z., Ma, Z., Zhang, Y., & He, B. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3361980 · Cited by 2
“It employs the NVIDIA Jetson Nano, an efficient embedded microprocessor, providing ample computational power for real-time MTL model execution with low power consumption [31].”
MethodTijerina, M. B., Piña‐Monarrez, M. R., Izaguirre, J. M. V., et al. (2024) · Revista de Ciencias Tecnológicas · doi:10.37636/recit.v7n3e350 · Cited by 1
Zheng, Q., Jin, J., Shen, Z., et al. (2024) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2024.3469545 · Cited by 1
“Moreover, AI technology such as deep learning prefers to leverage computationally expensive models to achieve high processing accuracy [7].”
BackgroundEnabling Memory-Augmented Neural Networks for Efficient Edge Applications
Zenozian, E. K., Kamal, M., Afzali‐Kusha, A., & Pedram, M. (2025) · Springer eBooks · doi:10.1007/978-3-031-71436-8_16 · Cited by 2
Optimizing Deep Learning Inference on RISC-V CPUs within the OpenVINO Toolkit
Sidorova, A., Mukhin, I., & Кустикова, В. Д. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-80457-1_6 · Cited by 2
LightCL: Compact Continual Learning with Low Memory Footprint For Edge Device
Wang, Z., Cheng, F., Ji, K., & Huang, B. (2025) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1145/3658617.3697784 · Cited by 3
Cheng, J.-A., Li, Q., Souri, A., Lei, X., Zhang, C., & Gao, M. (2025) · Cluster Computing · doi:10.1007/s10586-025-05226-y · Cited by 2
“Another line approach is to compress high-complexity crowd counting models through pruning and quantization, which reduces parameters and computational costs [47].”
Self-citationCoCoI: Distributed Coded Inference System for Straggler Mitigation
Liu, X., Huang, C. H., & Tang, M. (2025) · International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt) · doi:10.23919/wiopt66569.2025.11123371 · Cited by 1
“However, CNN inference is typically computationally intensive, while the computational resources of a single edge device are usually too limited to serve inference requests promptly.”
Federated Inference: Towards Collaborative and Privacy-Preserving Inference over Edge Devices
Fan, B., Su, X. J., Tarkoma, S., & Hui, P. (2025) · ACM SIGCOMM Conference Posters and Demos · doi:10.1145/3744969.3748418 · Cited by 4
“However, complex AI inference on resource-constrained edge devices (e.g., IoT boards and Raspberry Pi) remains challenging due to limited computation, memory, and energy budgets [7].”
Gao, H., Zhou, X., Lin, Y., Tian, M., Liu, L., & Shi, C. (2025) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/ae0a78 · Cited by 4
“They are thus not suitable for ubiquitous edge devices in the era of artificial intelligence-of-things, where massive sensory data call for in-situ and adaptive processing under stringent energy, cost and latency budgets [4].”
2023 Index Proceedings of the IEEE Vol.111
(2023) · Proceedings of the IEEE · doi:10.1109/jproc.2023.3345570
Tsalera, E., Stratogiannis, D. G., Papadakis, A. E., Voyiatzis, I., & Σαμαράκου, Μ. (2023) · Pan-Hellenic Conference on Progress in Computing and Informatics (PCI) · doi:10.1145/3635059.3635070
“A detailed survey in the area of efficient inference is provided in [3], where the main research challenges are identified and analyzed.”
BackgroundA comparative study of FPGA and CGRA technologies in hardware acceleration for deep learning
Luo, J. (2024) · Applied and Computational Engineering · doi:10.54254/2755-2721/51/20241212
Maximizing training efficiency through intelligent data exposure
Depoian, A. C., Bailey, C. P., & Guturu, P. (2024) · doi:10.1117/12.3013967
Context-Driven Device-Edge Collaborative Vision Transformer Models for Edge AiPreprint
Nimi, S. T., Arefeen, M. A., Uddin, M. Y. S., Debnath, B. K., & Chakradhar, S. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4862594
Optimized Deep Learning Classification Model for Intelligent Edge devices
Naveen, S., & Kounte, M. R. (2024) · Journal of Engineering Science and Technology Review · doi:10.25103/jestr.173.11
Edge-Optimized Model for Multimedia Classification using Linguistic Metadata
Bharitkar, S., & Páez, T. (2024) · IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) · doi:10.1109/icasspw62465.2024.10626175
“Transfer learning and optimization of these networks for computation or memory [26] can reduce the number of parameters, but can also reduce performance (e.g. [27]).”
BackgroundEESiamese: Energy-efficient Siamese Neural Network for Constrained Devices
Calisti, L., Contoli, C., Fabrizio, G. D., Kania, N., & Lattanzi, E. (2024) · International Conference on Machine Learning Technologies (ICMLT) · doi:10.1145/3674029.3674073
“Another strategy, often combined with compression, is to leverage hardware acceleration by gaining access to high-performance computing capabilities such as Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs) [13, 20, 37].”
Offload Strategy for Edge Computing in Satellite Networks Based on Software Defined Network
Jiang, Y., Liu, Z., Gui, Y., et al. (2024) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2024.057353
Ferreira, B. L. M., Sousa, A. J., Reis, L. P., de Sousa, A. A., Rodrigues, R., & Rossetti, R. J. F. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73503-5_5
Wang, C., Masuda, Y., & Ishihara, T. (2024) · IEICE Transactions on Electronics · doi:10.1587/transele.2024lhp0004
“With the explosive spread of edge AI in today’s advanced information society, energy-efficient AI accelerators are highly demanded [3].”
Path Finding Camera Network Using Edge Device
George, V., Rao, S., Singh, S., et al. (2024) · doi:10.1109/icec59683.2024.10837052
“Optimization of deep learning interpretations utilizing edge computation is well analyzed in [3] in which a comprehensive survey of transfer learning techniques in image classification is presented.”
Barhoush, M., Ayad, A., Kohankhaki, M., & Schmeink, A. (2024) · International Conference on Federated Learning Technologies and Applications (FLTA) · doi:10.1109/flta63145.2024.10839827
Collaborative Inference for Adaptive DNN Pipeline-Aware Alignment
Ren, J., Liu, L., Bi, R., Li, S., Li, S., & Lv, X. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-0055-7_25
Dishari: A Novel Gesture-Based Educational Application for Specially Challenged People
Bhattacharya, T., & Mohammad, I. A. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-85933-5_12
A Comparison of Quantization Techniques for Mobile Olive Disease Detection
Chouchf, S., Akhal, H. E., Zarrik, W., et al. (2025) · International Conference on Innovative Research in Applied Science, Engineering and Technology (IRASET) · doi:10.1109/iraset64571.2025.11008013
“These techniques are widely applicable in mobile and edge devices, particularly in scenarios requiring real-time processing, such as autonomous driving and video analysis [2] .”
InfluentialComplexity-Reduced End-to-End Fetal ECG Signal Recovery and QRS Complex Detection
Remus, J. C., & da Silveira, T. L. T. (2025) · doi:10.5753/sbcas.2025.7731
Aragonda, G., Najafi, D., Vungarala, D., et al. (2025) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3716368.3735277
Meng, Z., Shu, L., Li, Z., Lv, K., Lin, C., Xiao, L. X., & Hao, J. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11044224
Design of Cost-Effective Binary Neural Network's Accelerator with Ultra-Low Latency
Zheng, Y., Hu, H., & Xu, Y. (2024) · doi:10.1109/incc64392.2024.00018
“It shows great hardware friendliness to resource-constrained on-device [2] and edge-device computing, especially for inference on edge, like face identification, autonomous driving[3], in which the limitation is resource constraint.”
Depoian, A. C. (2025) · University of North Texas Digital Library (University of North Texas) · doi:10.12794/metadc2443141
An Experimental Study of DNN Online Split Learning on Edge Devices
Zang, B., Qu, Y., Dong, C., et al. (2025) · doi:10.1109/scecs65243.2025.11065718
“Notably, the ever-growing scale of deep neural networks (DNNs) have led to a surge in memory and computational requirements during their training process, posing an insurmountable obstacle for resource-limited edge devices [6] [7] .”
MG2GS: Optimizing Resource Efficiency for AI Training with Cross-MEC Job Scheduling
Gao, Z., Tian, Y., Hu, Y., Hu, J., Gong, X., & Wang, W. (2024) · IEEE International Conference on High Performance Computing and Communications; IEEE International Conference on Smart City; IEEE International Conference on Data Science and Systems (HPCC/SmartCity/DSS) · doi:10.1109/hpcc64274.2024.00215
Hu, X., Zeng, J., Dai, M., et al. (2025) · Journal of Food Composition and Analysis · doi:10.1016/j.jfca.2025.108161
Madhini, B., & Supraja, P. (2025) · Springer proceedings in earth and environmental sciences · doi:10.1007/978-3-031-93887-0_15 · Cited by 1
Chen, Q., Wang, H., Zhang, H., et al. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5349327
Many Hands Make Light Work: Accelerating Edge Inference via Multi-Client Collaborative Caching
Liang, W., Liu, J., Xu, H., Qiao, C., & Huang, L. (2025) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde65448.2025.00013 · Cited by 1
“These considerations have catalyzed a shift toward edge computing [4, 5], where inference processes migrate closer to data sources.”
MSDF-Based Hardware Accelerators for Energy-Efficient Neural Networks in Edge Computing Applications
Cherati, S. M., & Sousa, L. A. (2025) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi65124.2025.11130344
“In resource-constrained environments of edge computing applications, such as wearable devices and the Internet of Things (IoT) devices, traditional arithmetic units face challenges in meeting the required computational requirements and energy efficiency demands [3], [4].”
Kong, C., Chen, J., Meng, D., et al. (2025) · Scientific Reports · doi:10.1038/s41598-025-15351-8
Challenges and Opportunities in Federated Learning for the Internet of Everything (IoE)
Bansal, S., Kumar, P., raj, T. P., Mishra, S., & Goyal, P. (2025) · Studies in systems, decision and control · doi:10.1007/978-3-031-99270-4_7 · Cited by 1
Inference Acceleration and Optimization of AI Compiler on Heterogeneous Computing Platforms
Xu, M., Wang, W., Sun, Y. J., et al. (2025) · Smart innovation, systems and technologies · doi:10.1007/978-3-032-06372-4_24
Real-Time Video Streaming and Per-Frame Analysis with Qwen2.5-VL for Maritime Surveillance
Li, T., & Zhang, F. (2025) · International Conference on Artificial Intelligence, Big Data and Algorithms (CAIBDA) · doi:10.1109/caibda65784.2025.11183532
“Complex deep learning models (like large object detectors or VLMs) typically require GPU acceleration, limiting their use on small edge devices [4].”
MuhsnHasan, M., Sravanthi, J., Suresh, N. V. L. G., Dhote, V., Roy, J., & Yuvaraj, K. (2025) · International Conference on Metaverse and Current Trends in Computing (ICMCTC) · doi:10.1109/icmctc62214.2025.11196149 · Cited by 1
“Although the system relies on lightweight deep learning models tuned for edge hardware and good accuracy [11], it achieves higher performance than naive random sampling.”
Hu, H., Liu, Y., Liu, Y., et al. (2025) · Sensors · doi:10.3390/s25206464
“In recent years, with the rapid development of the Internet of Things (IoT) and edge computing, there has been an increasingly urgent demand for energy-efficient and low-latency computing in resource-constrained sensor nodes [1–4].”
CrowdMesh: A Dynamic Model Parallel Training System on Mobile Devices
Wang, Z., Guo, B., Liu, S., Yu, Z., & Zhang, D. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3623109
“Current DL deployment typically combines edge-based inference with cloud-based retraining [7], [8], but privacy concerns and transmission costs have spurred interest in edge training [9], [10], [11].”
Coded Robust Aggregation for Distributed Learning Under Byzantine Attacks
Li, C., Xiao, M., & Skoglund, M. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3624620 · Cited by 1
“Compared to training on a single device, DL leverages the computational resources of various edge devices, thereby increasing training e ffi ciency [3], [4].”
Hao-wu, L., Hou, S.-F., & Ruan, S.-J. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3627085
“SA architectures can generally be categorized as weight-stationary (WS) or output-stationary (OS), each with tradeo ff s in data movement and power e ffi ciency [7].”
A Survey on Microprocessors Optimization Techniques for AI Tasks on Edge Devices
AlMudaweb, Z., & Ramadan, A. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-00329-4_51
Neuromorphic computing for energy-efficient machine intelligence
Latif, S., Zafar, S., Ahmad, J. E., Khan, M. Z., Ullah, F., & Khattak, A. A. (2025) · Engineering Research Express · doi:10.1088/2631-8695/ae1d0c
Enhancing Knowledge Distillation via Mutual Information Guided Channels
Ali, U. A. M. E., Maniamfu, P., & Kameyama, K. (2025) · Communications in computer and information science · doi:10.1007/978-981-95-4094-5_29
Zirui, L., Gulzar, H., Nakadai, K., et al. (2025) · Springer eBooks · doi:10.1007/978-981-96-8935-4_5
Liu, K., Wait, W., & Li, Y. (2025) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas53549.2025.11244463
“However, deploying these models on edge devices, which have limitations in power and processing resources, presents a significant challenge [3].”
Parameterized Low-Overhead Data Conversion Architectures for Posit Arithmetic
Shekhawat, D., Pangoria, M., Sreejith, V., et al. (2025) · Conference proceedings · doi:10.1109/mwscas53549.2025.11244518
Madisa, M. (2025) · World Journal of Materials Science and Technology · doi:10.11648/j.wjmst.20250202.12
AI-Driven Cloud Resource Optimization for Smart Cities
Ghazal, T. M., Kommula, A. R., & Al-Dmour, N. A.-H. (2025) · International Conference on Business Analytics for Technology and Security (ICBATS) · doi:10.1109/icbats66542.2025.11258510
“Shuvo et al.[17] examine the problems inherent in deploying deep learning (DL) models in edge devices which often have limited resources like memory, power, and computational capacity.”
Liu, Y. (2025) · Internet Technology Letters · doi:10.1002/itl2.70198
Research on Large Model Fine-Tuning and Edge Deployment Mechanisms in Power Scenarios
Huang, Y., Yang, Y., Lin, J., Jiang, Z., & Li, R. (2025) · International Conference on Power System and Power Engineering (PSPE) · doi:10.1109/pspe66589.2025.11296397
“Traditional deep learning methods typically rely on large amounts of labeled data for training and lack the ability to deploy efficiently on resource-constrained devices [10].”
An Improved Single-Stage Object Detection Model and Its Application to Oil Seal Defect Detection
Chen, Y., Wu, Y., Wu, X., He, W., He, G., & Cai, X. (2025) · Electronics · doi:10.3390/electronics15010128
“However, the increasing complexity of these models leads to significant computational and memory costs, limiting their deployment for resource-constrained edge devices [3–5].”
PocketKD: Adaptive Confidence Distillation and Lightweight Student Networks for Embedded Systems
Gan, H. (2025) · International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP) · doi:10.1145/3778534.3778654
“However, most high-precision models, such as ResNet and ViT [2], typically require large amounts of computation and storage space, making it difficult to directly deploy them to resource-constrained edge devices [3], such as Rasp-berry Pi, ARM embedded boards, and low-power industrial control units.”
Tiny Machine Learning for Enhanced Edge Intelligence
Alozie, E., Imoize, A. L., Olagunju, H. I., Faruk, N., Garba, S., & Olatunji, A. P. (2025) · Wiley eBooks · doi:10.1002/9781394294572.ch9
An Improved Lightweight WDCNN for Smart Gear Edge Diagnosis
Wu, Q., Huang, W., & Ding, X. (2026) · Lecture notes in mechanical engineering · doi:10.1007/978-981-95-3650-4_24
Metadata
- Authors
- Shuvo, M. M. H., Islam, S. K., Cheng, J., & Morshed, B. I.
- Affiliations
- University of Missouri (US)
- Texas Tech University (US)
- Venue
- Proceedings of the IEEE
- Published
- 2023-01
- Volume
- 111
- Issue
- 1
- Pages
- 42-91
- ISSN
- 0018-9219 (ISSN-L)
- 1558-2256
- OpenAlex
- W4312910656
- Cited by
- 436
- References
- 491
- Citations by year
- 2026: 159 · 2025: 173 · 2024: 82 · 2023: 22
- FWCI
- 22.67
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Proceedings of the IEEE (publishedVersion, cc-by)
- ThinkTech (Texas Tech University) (submittedVersion, cc-by)
- ThinkTech (Texas Tech University) (submittedVersion)
- Funders
- College of Engineering, University of Missouri, Columbia, MO, USA
- 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
- deep learning inference, edge computing, resource-constrained devices, inference acceleration, algorithm-hardware co-design, neural network accelerators, model compression, Internet of Things, embedded systems, energy-efficient computing
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.
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
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.