Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review
Open accessThe DOI matches its Crossref record; 410 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 101–200 of 420
FedTrack: A Collaborative Target Tracking Framework Based on Adaptive Federated Learning
Pan, Y., Zhu, C., Luo, L., Liu, Y., & Cheng, Z. (2024) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2024.3395292 · Cited by 8
“However, privacy concerns may prevent these edge devices from sharing their local data or consume excessive bandwidth and computing resources of the learning nodes [7].”
BackgroundImplementation and analysis of custom instructions on RISC-V for Edge-AI applications
Ajay Kumar, M., Kumar, V., John, D., & Shreejith, S. (2024) · International Symposium on Highly Efficient Accelerators and Reconfigurable Technologies (HEART) · doi:10.1145/3665283.3665342 · Cited by 12
Efficient NPU–GPU scheduling for real-time deep learning inference on mobile devices
Yu, C., Wang, M., Chen, S., Wang, W., Fang, W. W., Chen, Y., & Xiong, N. N. (2025) · Journal of Real-Time Image Processing · doi:10.1007/s11554-025-01670-6 · Cited by 11
“However, the deployment of DNNs in mobile scenarios faces significant challenges, including the computation-intensive nature of DNNs, the energy constraint of mobile devices [1, 2], and the real-time demands of AI applications.”
InfluentialHarnessing Approximate Computing for Machine Learning
Shakibhamedan, S., Aminifar, A., Vassallo, L., & TaheriNejad, N. (2024) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi61997.2024.00110 · Cited by 10
“This process not only accelerates inference times but also reduces the energy consumed per operation, making it particularly advantageous for embedded and mobile applications where power efficiency is crucial [27].”
BackgroundCheng, Z., Xia, X., Wang, H., Liwang, M., Chen, N., Fan, X., & Wang, X. M. (2025) · IEEE Transactions on Services Computing · doi:10.1109/tsc.2025.3607117 · Cited by 11
“Furthermore, the growing disparity between the limited resources on mobile devices (MDs) and the computational demands of inference services presents another significant hurdle [3].”
Garai, S., & Samui, S. (2026) · Neurocomputing · doi:10.1016/j.neucom.2026.134028 · Cited by 19
A lightweight residual dilated temporal transformer block for ECG classification on edge devices
Gracy, G. A., & Pravin, S. C. (2026) · Scientific Reports · doi:10.1038/s41598-026-35531-4 · Cited by 14
Case study of WebAssembly Runtimes for AI Applications on the Edge
Khelifa, S. E., Bagaa, M., Messaoud, A. O., et al. (2024) · Global Information Infrastructure and Networking Symposium · doi:10.1109/giis59465.2024.10449907 · Cited by 5
“Hence, there are many ways to optimize neural networks, such as pruning, quantization, and knowledge distillation [24], also another way of optimization is using level 2 optimization or level 3 optimization; the last and most important factor is the batch size of images depending on the target…”
MethodEMI: Energy Management Meets Imputation in Wearable IoT Devices
Hussein, D., Yamin, N., & Bhat, G. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3448379 · Cited by 6
“A common goal of EM in wearable and IoT devices is achieving energy neutral operation (ENO), whereby the energy consumed in a given horizon (e.g., a day) is equal to the harvested energy [5], [7].”
BackgroundPatel, R. K. (2025) · Computational Biology and Chemistry · doi:10.1016/j.compbiolchem.2025.108620 · Cited by 5
Efficient Federated Learning using Random Pruning in Resource-Constrained Edge Intelligence Networks
Chen, C., Jiang, B., Liu, S., Li, C., Wu, C., & Yin, R. (2023) · Globecom · doi:10.1109/globecom54140.2023.10437051 · Cited by 5
“To address this issue, model compression has been widely considered to accelerate the training process recently [1], [2].”
BackgroundScaling in Deep and Shallow Learning Architectures
Koresh, E., Halevi, T., Meir, Y., Dilmoney, D., Dror, T., Gross, R. D., Tevet, O., Hodassman, S., & Kanter, I. (2024) · Physica A Statistical Mechanics and its Applications · doi:10.1016/j.physa.2024.129909 · Cited by 5
Katare, D., Leroux, S., Janssen, M., & Ding, A. Y. (2025) · Computing · doi:10.1007/s00607-025-01427-w · Cited by 4
“The efficiency of vision transformers can be categorized by their memory, parameter, and computational (processor) requirements [14, 15].”
LGVM-YOLOv8n: A Lightweight Apple Instance Segmentation Model for Standard Orchard Environments
Han, W., Li, T., Guo, Z., Wu, T., Huang, W., Feng, Q., & Chen, L. (2025) · Agriculture · doi:10.3390/agriculture15121238 · Cited by 7
Deep Learning on RISC-V Platforms at the Edge: A Perspective on the Hardware and Software Support
Agosta, G., Galimberti, A., & Zoni, D. (2025) · ACM Computing Surveys · doi:10.1145/3772277 · Cited by 15
“…and optimization techniques for deep neural networks (DNNs) [33], with emphasis on their suitability for hardware acceleration [42, 175], and provide taxonomies of hardware-software (HW/SW) co-design methods and accelerator-oriented optimizations for efficient edge inference [2, 80, 157].”
Xiao, Q., Yu, L., Zhai, X., Jiang, W., Jin, Y., Yuan, K., & Shi, W. (2025) · Signal Processing · doi:10.1016/j.sigpro.2025.110386 · Cited by 5
DNN Partitioning for Cooperative Inference in Edge Intelligence: Modeling, Solutions, Toolchains
Hao, Y., Ding, N., Xia, W., Ge, H., & Xu, L. (2026) · ACM Computing Surveys · doi:10.1145/3786145 · Cited by 14
“Early works focus on architectural classifications, distinguishing between device-server and device-device collaboration [35, 38].”
Thermal Stress Analysis via Infrared Imaging for Safety Monitoring in Biomedical Electronic Devices
Laganà, F., Pratticò, D., Calcagno, S., Carlo, D. D., Foresta, F. L., & Angiulli, G. (2025) · International Workshop on Biomedical Applications, Technologies and Sensors (BATS) · doi:10.1109/bats67559.2025.11336186 · Cited by 4
“The CNN architecture is optimized for low-latency inference, making it suitable for real-time deployment in both laboratory and clinical environments [24, 25].”
Deployment-Aware Compression of Large Language Models for Edge Intelligence: A System-Level Survey
Dritsas, E., & Trigka, M. (2026) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2026.3656155 · Cited by 17
“These benchmarks often assume static architectures and fixed operator semantics, neglecting the variability introduced by compiler optimizations, memory layouts, and quantization-aware execution [97].”
A comparative study of neural network pruning strategies for industrial applications
Zenoozi, A. D., Erhan, L., Liotta, A., & Cavallaro, L. (2025) · Frontiers in Computer Science · doi:10.3389/fcomp.2025.1563942 · Cited by 7
A lightweight optimization framework for real-time object detector on the embedded GPU platform
Zhu, J., Feng, H., Zhong, S., et al. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-04219-z · Cited by 4
“An approach is to construct the network structure using Optimized Convolution (OC) techniques [15], such as Depthwise Separable Convolution (DSC) and Group Convo-lution, to achieve low computational complexity and reduced Fig.”
HeterScale: A hierarchical task scheduling framework for intelligent edge collaboration in IIOT
Zhang, R., Liu, W., Zhu, Z., et al. (2025) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2025.103536 · Cited by 4
Unraveling the Potential of Heart Rate Variability for Type 1 Diabetes Prediction
Shuvo, M. M. H., Titirsha, T., & ISLAM, S. K. (2023) · International Workshop on Biomedical Applications, Technologies and Sensors (BATS) · doi:10.1109/bats59463.2023.10303132 · Cited by 6
“…(AB), decision tree (DT), k-nearest neighbor (k-NN), na¨ıve Bayes (NB), random forest (RF), XGBoost (XB), and multi-layer perceptron (MLP)) are implemented, and tuned with grid search and five-fold cross-validation [15], in automatically classifying diabetic and healthy subjects from HRV features.”
MethodSelf-citationHe, P., Wu, R., Liu, D., Dou, J., Hayat, K., Shang, D., Pan, J., & Lin, H. (2024) · Journal of Animal Science · doi:10.1093/jas/skae098 · Cited by 6
Optimizing Neural Network Inference in Edge Robotics by Harnessing FPGA Hardware Acceleration
Rao, K. H. (2024) · Journal of Electrical Systems · doi:10.52783/jes.2529 · Cited by 4
“Several research efforts have demonstrated the effectiveness of FPGA-based acceleration in enhancing the performance of neural network inference tasks on edge devices [10].”
BackgroundKumar, S., Ranjan, V., Chakrabarti, A., Das, T. K., & Singh, A. (2024) · Expert Systems · doi:10.1111/exsy.13760 · Cited by 5
“Edge devices, characterised by limited memory, computational power, and energy resources, are often unable to support the high demands of large LLMs without substantial optimization (Shuvo et al. 2023; Lu et al. 2023).”
Detecting Numerical Deviations in Deep Learning Models Introduced by the TVM Compiler
Xia, Z.-C., Chen, Y., Nie, P., & Wang, Z. (2024) · Proceedings/Proceedings - International Symposium on Software Reliability Engineering · doi:10.1109/issre62328.2024.00018 · Cited by 6
Rangsikunpum, A., Amiri, S., & Ost, L. C. (2024) · IET Computers & Digital Techniques · doi:10.1049/cdt2/6214436 · Cited by 4
YOLO_SSP: An Auto-Algorithm to Detect Mature Soybean Stem Nodes Based on Keypoint Detection
Wu, Q., Liu, H., Zhu, H., Wang, C., Wang, H., Han, Z., Zhao, L., & Liu, F. (2025) · Agronomy · doi:10.3390/agronomy15051128 · Cited by 7
Advances in AI and ML for Cloud Computing: A Review of Algorithms, Challenges, and Innovations
Ramamoorthi, V. (2025) · International Journal of Scientific Research in Science and Technology · doi:10.32628/ijsrst2513120 · Cited by 4
Shuvo, M. M. H., Titirsha, T., Hossain, K. S. M. T., Gonzalez, G. L., & ISLAM, S. K. (2024) · IEEE International Symposium on Medical Measurements and Applications (MeMeA) · doi:10.1109/memea60663.2024.10596811 · Cited by 5
“Benefiting from a compact architecture, the model achieves efficient computation, opening possibilities for edge applications in diabetes management [23]–[25].”
Self-citationTraffic Density Estimation by Distributed Proxy Model Learning for Internet of Vehicle
Li, Q., Cheng, J., Gao, M., Chen, J., & Jeon, G. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3542534 · Cited by 4
“Nevertheless, current deep learning models [4], [5], [6] often possess a large number of parameters, which makes them less suitable for deployment on resource-constrained edge devices [7].”
Self-citationSV-SAE: Layer-Wise Pruning for Autoencoder Based on Link Contributions
Rheey, J., & Park, H. (2025) · IEEE Access · doi:10.1109/access.2025.3565296 · Cited by 3
“In many practical applications, training, and inference are separated to address the computational limitations of edge devices [2].”
Dai, X., Cheng, G., Yang, L., Wang, Y., Li, Z.-H., Han, S., & Liu, J. (2025) · Agriculture · doi:10.3390/agriculture15151643 · Cited by 2
Zhao, R., Luo, C., Zhang, Q., Wei, Q., Zhang, X., Wang, M., Zhang, J., & Li, H. (2025) · Smart Agricultural Technology · doi:10.1016/j.atech.2025.101330 · Cited by 6
NExNet Seg: Neuron Expansion Network for Medical Image Segmentation
Reyes-Angulo, A. A., & Paheding, S. (2025) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw67362.2025.00005 · Cited by 3
“These advancements, however, often come with a high computational cost due to the large number of trainable parameters—e.g., U-Net contains approximately 31 million parameters [26]—posing challenges for deployment in resource-constrained clinical environments [27].”
Hu, S., Lu, Z., Xu, X., Deng, R., Du, X., & Duan, Q. (2025) · Journal of Industrial Information Integration · doi:10.1016/j.jii.2025.100955 · Cited by 18
LSTM Hardware Inference Accelerator for LiteRT
Mannes, G., Manor, E., & Greenberg, S. (2025) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2025.3609633 · Cited by 6
“ML inference at the edge, particularly within low-power MCUs, is gaining increased interest amongst the industrial and academic ML community [1].”
Cognitive Internet of Things: A Review of Theory, Applications, and Recent Advances
Giuliano, A., McCafferty-Leroux, A., Yawney, J., & Gadsden, S. A. (2025) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2025.3615461 · Cited by 15
“As TinyML continues to evolve, the synergy between model compression, hardware-aware optimizations, and emerging AI accelerators will drive the next generation of ultra-low-power deep learning applications [119].”
InfluentialHuang, C., Chen, Y., Zhou, B., Cai, X., Zhai, Z., Zhang, J., & Zhan, Y. (2025) · Sensors · doi:10.3390/s25247529 · Cited by 6
“However, the computational and storage requirements of most deep models substantially exceed the capacity of typical edge hardware, prompting the development of model compression and lightweight design techniques [37].”
Tong, Y., Yuan, J., & Hu, C. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3651729 · Cited by 4
“These applications must run under stringent compute and battery limits [3], [4].”
Shi, Z., Zhang, H., Lin, Y., Cheng, J., Wang, L., Luan, F., Chang, X., & Hu, Z. (2026) · Trends in Food Science & Technology · doi:10.1016/j.tifs.2026.105850 · Cited by 4
Self-citationGhavami, B., Sajadi, M., Shannon, L., & Wilton, S. J. E. (2024) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi61997.2024.00049 · Cited by 5
“FPGAs, being flexible and reprogrammable devices, offer customization options to deliver high performance tailored to specific DNN applications [4] [5].”
BackgroundPipeMCTS: Pipeline Inference Optimization for Edge Computing via Surrogate Model-Guided MCTS
Wang, Z., Ding, J., Xian, D., Jin, Y., Hua, D., Hua, W., Lv, W., & Lin, C. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3546577 · Cited by 3
Peng, Y., & Li, E. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.128427 · Cited by 3
Hardware-Accelerated Edge AI Orchestration on the Multi-Tier Edge-to-Cloud Continuum
Palomares, J., Coronado, E., Tzenetopoulos, A., et al. (2025) · Journal of Network and Systems Management · doi:10.1007/s10922-025-09959-4 · Cited by 3
“For instance, the analysis presented in [55] demonstrates that deploying AI inference on GPUs can reduce latency by up to 10 × while also achieving substantial energy savings.”
Inés, A., Diaz‐Pinto, A., Domínguez, C., Heras, J., Mata, E. J., & Pascual, V. (2023) · Soft Computing · doi:10.1007/s00500-023-09109-5 · Cited by 3
Performance Modelling-Driven Optimization of RISC-V Hardware for Efficient SpMV
Rodrigues, A., Sousa, L. A., & Ilić, A. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-40843-4_36 · Cited by 3
Unstructured Pruning for Multi-Layer Perceptrons with Tanh Activation
Kalyanam, L. K., & Katkoori, S. (2023) · IEEE International Symposium on Smart Electronic Systems (iSES) · doi:10.1109/ises58672.2023.00025 · Cited by 3
“The removal of neurons or weights, often driven by heuristic criteria, can lead to an unforeseen degradation in model performance [4], [5].”
BackgroundWu, Q., Zhang, Y., Yang, C., & Sun, J. Y. (2023) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads60453.2023.00295 · Cited by 3
Adaptive knowledge transfer for data-free low-bit quantization via tiered collaborative learning
Yang, Z., Lin, T., Li, C., Qian, B., Yang, X., & Wei, X. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130097 · Cited by 3
GADF-FIRViT: A Lightweight Interference Classifier for Infrared Attitude Measurement
Xu, M., Han, X., Sun, Q., Chen, X., Bu, X., & Cao, Y. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3581658 · Cited by 4
XtremeLLMs: Towards Extremely Large Language ModelsPreprint
Mienye, I. D., Swart, T. G., & Obaido, G. (2024) · Preprints.org · doi:10.20944/preprints202408.1483.v1 · Cited by 3
Lightweight Design and Optimization Methods for Dcnns: Progress and FuturesPreprint
Long, H., Bi, W., & Sun, J. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4972503 · Cited by 9
Time-series visual representations for sleep stages classification
Ederli, R. P., Oliveros, D. A. V., Soriano-Vargas, A., Rocha, A. D. R., & Dias, Z. (2025) · PLoS ONE · doi:10.1371/journal.pone.0323689 · Cited by 3
“Deploying Deep Learning technologies in wearable devices is challenging due to their computational and energy limitations [47].”
Zhu, M., Liu, J., Lu, Y., Hu, Z., Li, Y., & Zhou, K. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113882 · Cited by 5
Deep Learning-Based Approach to Automated Monitoring of Defects and Soiling on Solar Panels
Hamdi, A., Noura, H., & Azar, J. (2025) · Future Internet · doi:10.3390/fi17100433 · Cited by 5
“The demand for real-time analysis for proactive maintenance [17] further imposes strict efficiency constraints, especially on edge devices [18].”
Ling, T., Singh, V., Qian, C., Bießmann, F., & Schiele, G. (2025) · IEEE International Smart Cities Conference (ISC2) · doi:10.1109/isc266238.2025.11293267 · Cited by 5
“Lightweight architectures such as LSTMs and Transformers have been widely adopted to meet these constraints while preserving time-series modeling capacity [16].”
Zhou, H., Su, Y., Sun, Y., et al. (2025) · Information Processing in Agriculture · doi:10.1016/j.inpa.2025.12.006 · Cited by 2
Patriananda, T., Istiqomah, & Suratman, F. Y. (2026) · IEEE Access · doi:10.1109/access.2026.3663174 · Cited by 2
“Second, the model is optimized through quantization , pruning , knowledge distillation , and mixed precision , and then executed using efficient runtime /tooling (e.g., TensorFlow Lite, with model interoperability via Open Neural Network Exchange (ONNX)) [32], [33].”
Le, Q. P., Vo, T. T., Ha, D., Vi, L. T. A., Ahn, J. S., Lee, B.-I., Choi, J., & Oh, J. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3666729 · Cited by 5
“While inference ultimately runs on edge devices, training DL models directly on the MCU is di ffi cult [45].”
Gao, H., Liu, Y., He, Z., Sun, Y., Wang, X., & Hu, C. (2026) · European Journal of Remote Sensing · doi:10.1080/22797254.2026.2642998 · Cited by 2
A Multi-Scale Dense Perception and Scale-Adaptive Approach for Blueberry Ripeness Detection
Guo, S., Yang, N., & Pang, S. (2026) · Foods · doi:10.3390/foods15071161 · Cited by 2
“Zhang et al. validated the feasibility of detecting blueberry ripeness using different versions of YOLOv11, with the YOLOv11m model achieving detection accuracies of 90% for ripe blueberries and 81% for unripe blueberries [28].”
Abetu, M. T., Urgessa, T., & Tune, K. K. (2026) · Smart Agricultural Technology · doi:10.1016/j.atech.2026.102091 · Cited by 5
Khan, F. M. A., Hallaq, M., Abou-Zeid, H., Erak, O., Waqar, O., Hassan, S. A., Alhussein, O., & Hossain, E. (2026) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2026.3682638 · Cited by 8
“The increasing complexity of DL models has intensified the need for efficient deployment on resource-constrained devices such as mobile phones, embedded systems, and edge devices [13].”
Kumar, S., Malviya, R., Sridhar, S. B., Wadhwa, T., & Shareef, J. (2026) · Journal of Alloys and Compounds Communications · doi:10.1016/j.jacomc.2026.100200 · Cited by 2
DNN Memory Footprint Reduction via Post-Training Intra-Layer Multi-Precision Quantization
Ghavami, B., Kamjoo, A., Shannon, L., & Wilton, S. J. E. (2024) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed60706.2024.10528676 · Cited by 3
An effective two-stage channel pruning method based on two-dimensional information entropy
Xu, Y., Yang, J., Wang, R., & Li, H. (2024) · Applied Intelligence · doi:10.1007/s10489-024-05615-7 · Cited by 2
“…point operations(FLOPs), difficulty in deployment has severely hampered the development and application of DNN. Deep neural network compression [9, 10] has been considered an effective method to solve this problem, which aims to reduce as many redundant parameters as possible without significantly…”
Liu, Y., Andhare, A., & Kang, K.-D. (2024) · Italian National Conference on Sensors · doi:10.3390/s24165262 · Cited by 2
Reliability Analysis of Phase Change Memory-Based Neuromorphic Circuits
Titirsha, T., Shuvo, M. M. H., & ISLAM, S. K. (2024) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi61997.2024.00144 · Cited by 2
“Although CPU performance and clock rates have improved, the growing gap between computational capability and memory bandwidth reduces system efficiency [2].”
BackgroundSelf-citationTang, M., Liu, N., Yang, T., Fang, H., Qiu, L., Tan, Y., Chen, X., Liu, D., Zhong, K., & Ren, A. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3443694 · Cited by 4
“However, the explosively increased model size hinders the deployment of deep learning models on embedded devices, which have limited computational and storage resources [4], [5], [6].”
BackgroundProposing YOLOv5 Model on Jetson Nano and Evaluating Map Index
Anh, Q. D., Quoc, H. L., Bao, C. N. L., et al. (2024) · IEEE International Symposium on Telecommunication Technologies · doi:10.1109/istt63363.2024.10750644 · Cited by 2
Sainz-Estebanez, U., Martinez-Corral, U., & Basterretxea, K. (2024) · Conference on Design of Circuits and Integrated Systems · doi:10.1109/dcis62603.2024.10769177 · Cited by 2
“Fully performing the growing number of AI inferences in the cloud is not only undesirable, it is unsustainable [1].”
Ling, T., Qian, C., Klann, T. M., Hoever, J., Einhaus, L., & Schiele, G. (2024) · Sensors · doi:10.3390/s25010083 · Cited by 6
“However, deploying models on resource-constrained hardware requires reducing numerical precision to minimize memory usage and accelerate inference [22].”
Enhancing μNAS for 1D CNNs on Microcontrollers
Capello, A., Berta, R., Ballout, H., et al. (2025) · Lecture notes in electrical engineering · doi:10.1007/978-3-031-71518-1_59 · Cited by 2
FPGA-Accelerated Semantic Segmentation for Urban Scenes
Shen, T., Zuo, Y., Zheng, H., Zhang, L., Hu, C., & Liu, H. (2024) · International Conference on Machine Vision, Image Processing and Imaging Technology (MVIPIT) · doi:10.1109/mvipit65697.2024.00014 · Cited by 2
Reducing Errors and Powers in LPDDR for DNN Inference: A Compression and IECC-Based Approach
Hong, J.-Y., Jang, J.-W., Cho, S.-H., Kong, Y., Kim, S., Kang, Y., Ko, J., Chung, J., & Yang, J.-S. (2025) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2025.103409 · Cited by 3
Wang, X., Li, G., & Liu, F. (2025) · Electronics · doi:10.3390/electronics14102038 · Cited by 4
Hu, X., Geng, X., Mao, Z., Han, J., & Jiang, H. (2025) · Design, Automation, and Test in Europe · doi:10.23919/date64628.2025.10993198 · Cited by 3
“MAC designs have been tailored to both floating-point and fixed-point computations [2]–[4].”
Wang, C., & Liu, H. (2025) · Symmetry · doi:10.3390/sym17060891 · Cited by 3
Early Termination of the MSDF Computations towards Efficient Inference in Neural Networks
Cherati, S. M., Barzegar, M., & Sousa, L. A. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11043511 · Cited by 2
An Improved Multi-task Model for Instance Segmentation and Pose Estimation Based on YOLOv11
Jiang, Y., Dai, Z., Chen, Y., & Chu, Z. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9856-1_19 · Cited by 2
Unleashing AI-Empowered Slices on Mobile Networks for Natively Cognitive Service Delivery
Moreira, R., Moreira, L. F. R., & Silva, F. D. O. (2025) · IEEE Access · doi:10.1109/access.2025.3599356 · Cited by 9
“The integration of AI capabilities closer to user devices has received increasing attention, consequently increasing the demand for processing and generating insights from the data volume [7].”
Lata, K., Singh, P., Saini, S., & Cenkeramaddi, L. R. (2025) · Computer Methods and Programs in Biomedicine Update · doi:10.1016/j.cmpbup.2025.100205 · Cited by 3
Stojanovic, D. H., Sentic, S., Stojanovic, N. M., & Stamenkovic, T. (2025) · Computer Science and Information Systems · doi:10.2298/csis240503020s · Cited by 3
“Shuvo et al. in [18] address the challenges of deploying DNNs on edge devices.”
Efficient Pipelining of Synchronous Dataflow Graphs Via Graph Conversion
Ma, M., Hou, J., Xiang, D., et al. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3355190 · Cited by 1
“These applications are widely used in fields, such as signal processing, multimedia, high-performance computing, and artificial intelligence [1], [2], [3], [4].”
Lightweight and Low-Parametric Network for Hardware Inference of Obstructive Sleep Apnea
Paul, T., Hassan, O., McCrae, C. S., ISLAM, S. K., & Mosa, A. S. M. (2024) · Diagnostics · doi:10.3390/diagnostics14222505 · Cited by 1
“Moreover, a large AI model will require significant amounts of memory and storage for storing network parameters and intermediate data [37].”
Self-citationEfficient recurrent real video restoration
Buades, A., & Lisani, J.-L. (2024) · Digital Signal Processing · doi:10.1016/j.dsp.2024.104851 · Cited by 1
A Feasibility Study on Containerization of Traditional Embedded Systems
Wang, W., & Li, J. (2024) · International Conference on Electronic Information Engineering and Computer Communication (EIECC) · doi:10.1109/eiecc64539.2024.10929091 · Cited by 1
“No matter industrial control jobs or deep learning model training can both be deployed on the device clusters [3].”
Sensor-Guided Adaptive Machine Learning on Resource-Constrained Devices
Papst, F., Kraus, D., Rechberger, M., & Saukh, O. (2024) · International Conference on the Internet of Things (IoT) · doi:10.1145/3703790.3703801 · Cited by 3
“DA, on the other hand, are robust towards noisy sensor data, as they don’t rely on sensor guidance.”
Low-Memory-Footprint CNN-Based Biomedical Signal Processing for Wearable Devices
Kokhazad, Z., Gkountelos, D., Kokhazadeh, M., Bournas, C., Κεραμίδας, Γ., & Kelefouras, V. I. (2025) · IoT · doi:10.3390/iot6020029 · Cited by 2
“If we consider that wearable devices (e.g., wristbands and smartwatches) are typically devices with limited memory and computing resources, the execution of DNN models in these devices is a challenging task [14].”
Challenges and Future Opportunities of Geospatial Modeling for Natural Resource Assessment
Yasin, E. H. E., Czimber, K., Gadallah, N. A. H., et al. (2025) · Advances in geographic information science · doi:10.1007/978-3-031-84698-4_22 · Cited by 1
Axelsson, K., Sheikhsarraf, M., Kutter, C., & Richter, M. (2025) · Sensors · doi:10.3390/s25123784 · Cited by 2
“Nowadays, artificial intelligence (AI) models are becoming more efficient and can be implemented in limited embedded systems [24–27].”
Wu, Q., Wang, Z., Xu, Y., Huang, W., & Ding, X. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3583820 · Cited by 1
“For model lightweighting methods, pruning, quantization, and knowledge distillation [27] can effectively achieve model compression.”
Challenges and Benchmarking of Object Detection Models on Edge AI SoCs
Kong, C. S., Mo, P., Zeng, T., Luo, Q., & Mao, R. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9908-7_21 · Cited by 2
Variational inference-aided neural architecture search for secure deep learning implementation
Yadav, A. K., Gupta, V. K., Bhati, G. S., & Kumar, B. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.131163 · Cited by 2
DGPAS: DQN-GRU guided distributed DNN pipeline training and adjacent scheduling in edge networks
Li, J., Wang, X., Chen, H., Wu, Z., Zhu, Z., Cao, J., & Buyya, R. (2025) · Computer Networks · doi:10.1016/j.comnet.2025.111592 · Cited by 3
From Circuits to Cloud: An Experience Report of a Full Stack IoT Course Curriculum
Louis, J. D., Morrow, N. S. J., & Gilbert, J. E. (2025) · ASEE Annual Conference & Exposition · doi:10.18260/1-2--56604 · Cited by 1
A review of parallel processing in resource-constrained Internet of Things (IoT) devices
Albalawi, N. S., Alanazi, A. G., Alshammari, S. A., Alhamazani, F., & Alshammari, A. A. (2025) · International Journal of ADVANCED AND APPLIED SCIENCES · doi:10.21833/ijaas.2025.09.003 · Cited by 2
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.