FANN-on-MCU: An Open-Source Toolkit for Energy-Efficient Neural Network Inference at the Edge of the Internet of Things
Open accessThe DOI matches its Crossref record; 58 of the 59 references it cites were found in a registry or an index.
Abstract
The growing number of low-power smart devices in the Internet of Things is coupled with the concept of “edge computing” that is moving some of the intelligence, especially machine learning, toward the edge of the network. Enabling machine learning algorithms to run on resource-constrained hardware, typically on low-power smart devices, is challenging in terms of hardware (optimized and energy-efficient integrated circuits), algorithmic, and firmware implementations. This article presents a FANN-on-MCU, an open-source toolkit built upon the fast artificial neural network (FANN) library to run lightweight and energy-efficient neural networks on microcontrollers based on both the ARM Cortex-M series and the novel RISC-V-based parallel ultralow-power (PULP) platform. The toolkit takes multilayer perceptrons trained with FANN and generates code targeted to low-power microcontrollers. This article also presents detailed analyses of energy efficiency across the different cores, and the optimizations to handle different network sizes. Moreover, it provides a detailed analysis of parallel speedups and degradations due to parallelization overhead and memory transfers. Further evaluations include experimental results for three different applications using a self-sustainable wearable multisensor bracelet. The experimental results show a measured latency in the order of only a few microseconds and power consumption of a few milliwatts while keeping the memory requirements below the limitations of the targeted microcontrollers. In particular, the parallel implementation on the octa-core RISC-V platform reaches a speedup of 22× and a 69% reduction in energy consumption with respect to a single-core implementation on Cortex-M4 for continuous real-time classification.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The text we found for this paper looks like another document, so the self-check wasn’t run.
Citations
Cites
Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.123 · Cited by 19,741 · cited 2×
“Recent results on machine learning models demonstrate impressive classification accuracy, in some cases even outperforming humans [1].”
BackgroundInternet of Things for Smart Cities
Zanella, A., Bui, N., Castellani, A., Vangelista, L., & Zorzi, M. (2014) · IEEE Internet of Things Journal · doi:10.1109/jiot.2014.2306328 · Cited by 6,292 · cited 1×
“One essential feature of machine learning and in particular neural networks (NNs) is their flexibility that makes them suitable for a wide range of applications, including computer vision [2], [3], natural language processing [4], [5], biomedical [6], [7], and several others [8]–[11].”
BackgroundEdge Computing for the Internet of Things: A Case Study
Premsankar, G., Di Francesco, M., & Taleb, T. (2018) · IEEE Internet of Things Journal · doi:10.1109/jiot.2018.2805263 · Cited by 697 · cited 1×
“The reasons are the massive workload on the IoT network, the cost of the communication infrastructure, the required energy for data transmission, and, more generally, reliability, latency, and privacy concerns [15], [16].”
BackgroundNear-Threshold RISC-V Core With DSP Extensions for Scalable IoT Endpoint Devices
Gautschi, M., Schiavone, P. D., Traber, A., Loi, I., Pullini, A., Rossi, D., Flamand, E., Gurkaynak, F. K., & Benini, L. (2017) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2017.2654506 · Cited by 507
BLE Beacons for Internet of Things Applications: Survey, Challenges, and Opportunities
Jeon, K. E., She, J., Soonsawad, P., & Ng, P. C. (2018) · IEEE Internet of Things Journal · doi:10.1109/jiot.2017.2788449 · Cited by 391 · cited 1×
“Thus, bringing intelligence to the edge is creating fascinating challenges for industrial and academic researchers [10], [12], [21].”
BackgroundCan, Y. S., Chalabianloo, N., Ekiz, D., & Ersoy, C. (2019) · Sensors · doi:10.3390/s19081849 · Cited by 349 · cited 1×
“Many studies identified MLPs to be the best or one of the best algorithms to solve tasks in the IoT domain using wearable devices [57]–[60].”
BackgroundCMSIS-NN: Efficient Neural Network Kernels for Arm Cortex-M CPUs
Oliveira, A., Costa, D., Moreira, G., Martins, J., & Pinto, S. (2018) · arXiv (Cornell University) · doi:10.4230/oasics.ng-res.2025.4 · Cited by 288 · cited 1×
“In parallel, ARM has been developing CMSIS-NN [48], an additional library for their Cortex Microcontroller Software Interface Standard (CMSIS), in order to provide optimal-performance implementations of layers commonly present in deep NNs with speedups in the order of 4.6 × over a simple baseline…”
MethodInfluentialA Robust and Energy-Efficient Classifier Using Brain-Inspired Hyperdimensional Computing
Rahimi, A., Kanerva, P., & Rabaey, J. M. (2016) · Proceedings of the 2016 International Symposium on Low Power Electronics and Design · doi:10.1145/2934583.2934624 · Cited by 276 · cited 1×
“Today’s IoT devices host microcontrollers (MCUs), especially from the ARM Cortex-M family, which are able to achieve power consumption in the order of mW and computational resources in the order of hundreds of MOPS [1], [26], [27].”
BackgroundYodaNN: An Architecture for Ultralow Power Binary-Weight CNN Acceleration
Andri, R., Cavigelli, L., Rossi, D., & Benini, L. (2018) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2017.2682138 · Cited by 247 · cited 1×
“The most promising approaches to improve the performance of ultra-low-power processors are parallelism, low-power fixed-function hardware accelerators [32], [33], memory access optimization [34], and near-threshold technology [35], [36].”
BackgroundSelf-citationFrom Cloud Down to Things: An Overview of Machine Learning in Internet of Things
Samie, F., Bauer, L., & Henkel, J. (2019) · IEEE Internet of Things Journal · doi:10.1109/jiot.2019.2893866 · Cited by 240 · cited 1×
“to the sensors where the data are collected [13].”
Background
Research works printed but not linked: 3
- 2019, MYO Gesture Control Armband—Wearable Technology by Thalmic Labs
- 2017, CMSIS-Cortex Microcontroller Software Interface Standard
- kuyumcu, 2007, FANN Tool
Cited by
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey
Wang, X., Han, Y., Leung, V. C. M., Niyato, D., Yan, X., & Chen, X. (2020) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2020.2970550 · Cited by 1,498
Self-citationA review on TinyML: State-of-the-art and prospects
Ray, P. P. (2022) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2021.11.019 · Cited by 460
Ingolfsson, T. M., Hersche, M., Wang, X., Kobayashi, N., Cavigelli, L., & Benini, L. (2020) · 2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC) · doi:10.1109/smc42975.2020.9283028 · Cited by 353
“However, accurate networks such as the TPCT model have 7.78 M train-able parameters and require 1.73 billion multiply-accumulate (MAC) operations per inference, which is out of reach of a typical low-power MCU with few MB of Flash and few hundreds of kB of RAM [18].”
BackgroundInfluentialSelf-citationKhalil, R. A., Saeed, N., Masood, M., Fard, Y. M., Alouini, M.-S., & Al-Naffouri, T. Y. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3051414 · Cited by 318
Edge Computing on IoT for Machine Signal Processing and Fault Diagnosis: A Review
Lu, S., Lu, J., An, K., Wang, X., & He, Q. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3239944 · Cited by 295
“One of the main reasons is that edge computing is highly depended on both hardware and algorithm, and the implementation of fault diagnosis methods on the specific processors is still a difficulty and challenge [7].”
BackgroundSelf-citationTinyML-Enabled Frugal Smart Objects: Challenges and Opportunities
Sanchez-Iborra, R., & Skarmeta, A. F. (2020) · IEEE Circuits and Systems Magazine · doi:10.1109/mcas.2020.3005467 · Cited by 295
“In turn, FANN-onMCU [55] is another open-source framework which is built upon the Fast Artificial Neural Network (FANN) library [63] aiming at running lightweight NNs on microcontrollers.”
MethodDeep Learning in Smart Grid Technology: A Review of Recent Advancements and Future Prospects
Massaoudi, M., Abu-Rub, H., Refaat, S. S., Chihi, I., & Oueslati, F. S. (2021) · IEEE Access · doi:10.1109/access.2021.3071269 · Cited by 256
Machine Learning for Microcontroller-Class Hardware: A Review
Saha, S. S., Sandha, S. S., & Srivastava, M. (2022) · IEEE Sensors Journal · doi:10.1109/jsen.2022.3210773 · Cited by 253
TinyML for Ultra-Low Power AI and Large Scale IoT Deployments: A Systematic Review
Schizas, N., Karras, A., Karras, C., & Sioutas, S. (2022) · Future Internet · doi:10.3390/fi14120363 · Cited by 215
Quantization and Deployment of Deep Neural Networks on Microcontrollers
Novac, P.-E., Boukli Hacene, G., Pegatoquet, A., Miramond, B., & Gripon, V. (2021) · Sensors · doi:10.3390/s21092984 · Cited by 210
Bruce Lee-Inspired Fluid Antenna System: Six Research Topics and the Potentials for 6G
Wong, K., Tong, K., Shen, Y., Chen, Y., & Zhang, Y. (2022) · Frontiers in Communications and Networks · doi:10.3389/frcmn.2022.853416 · Cited by 201
Green IoT: A Review and Future Research Directions
Alsharif, M. H., Jahid, A., Kelechi, A. H., & Kannadasan, R. (2023) · Symmetry · doi:10.3390/sym15030757 · Cited by 172
“One promising area of research in the field of MCUs and ICs is the use of ML algorithms to optimize their performance and reduce energy consumption [19].”
BackgroundAn Accurate EEGNet-based Motor-Imagery Brain–Computer Interface for Low-Power Edge Computing
Wang, X., Hersche, M., Tömekçe, B., Kaya, B., Magno, M., & Benini, L. (2020) · IEEE International Symposium on Medical Measurements and Applications (MeMeA) · doi:10.1109/memea49120.2020.9137134 · Cited by 148
“An increasing number of battery-operated wearable solutions, using microcontroller units (MCUs), are proposed to bring computing capabilities towards the “edge” to perform real-time near-sensor computation [10], [12], [13], [14].”
BackgroundSelf-citationTinyML: Tools, applications, challenges, and future research directions
Kallimani, R., Pai, K., Raghuwanshi, P., Iyer, S., & López, O. L. A. (2023) · Multimedia Tools and Applications · doi:10.1007/s11042-023-16740-9 · Cited by 185
“An open-source toolkit namely, fast artificial neural network (FANN)-on-MCU, which runs on PULP, is developed for reducing energy consumption by [35].”
A Guideline on Pseudorandom Number Generation (PRNG) in the IoT
Kietzmann, P., Schmidt, T. C., & Wählisch, M. (2021) · ACM Computing Surveys · doi:10.1145/3453159 · Cited by 129
“[118], who analyzed neural network inference on constrained IoT devices.”
BackgroundUnlocking Edge Intelligence Through Tiny Machine Learning (TinyML)
Zaidi, S. A. R., Hayajneh, A. M., Hafeez, M., & Ahmed, Q. Z. (2022) · IEEE Access · doi:10.1109/access.2022.3207200 · Cited by 129
“Typically local com-56 putation is several-fold less power-hungry than transmission over wireless channel [1].”
BackgroundLeveraging IoT-Aware Technologies and AI Techniques for Real-Time Critical Healthcare Applications
Shumba, A. T., Montanaro, T., Sergi, I., Fachechi, L., Vittorio, M. D., & Patrono, L. (2022) · Sensors · doi:10.3390/s22197675 · Cited by 142
López, O. L. A., Rosabal, O. M., Ruiz-Guirola, D. E., Raghuwanshi, P., Mikhaylov, K., Lovén, L., & Iyer, S. (2023) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2023.3323832 · Cited by 131
“For instance, authors in [324] adopted PULP-based MCU to facilitate the parallel run in non-neural ML algorithms, while a fast NN (FANN)-on-MCU framework supporting both ARM Cortex-M series MCUs and PULP-based MCUs is used in [320].”
MethodDiab, M. S., & Rodriguez–Villegas, E. (2022) · IEEE Access · doi:10.1109/access.2022.3206782 · Cited by 102
“The system was tested in an application for 870 stress detection based on collected readings from ECG and 871 GSR, implementing a multi-layer perceptron (MLP) using 872 fast artificial neural network (FANN) library [76].”
MethodAn Evolving TinyML Compression Algorithm for IoT Environments Based on Data Eccentricity
Signoretti, G., Silva, M., Andrade, P., Silva, I., Sisinni, E., & Ferrari, P. (2021) · Sensors · doi:10.3390/s21124153 · Cited by 73
“In this scenario, data processing is pushed to the “edge” of the network, even to IoT devices and sensors used primarily for data acquisition [27–29].”
MethodEnabling Resource-Efficient AIoT System With Cross-Level Optimization: A Survey
Liu, S., Guo, B., Fang, C., Wang, Z., Luo, S., Zhou, Z., & Yu, Z. (2023) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2023.3319952 · Cited by 83
“Prior efforts in cross-level co-design mainly include algorithm-hardware [339]–[341], compiler-hardware [342], [343], and algorithm-system [12], [344], [345] co-design.”
A TinyML Soft-Sensor Approach for Low-Cost Detection and Monitoring of Vehicular Emissions
Andrade, P., Silva, I., Silva, M., Flores, T., Cassiano, J., & Costa, D. G. (2022) · Sensors · doi:10.3390/s22103838 · Cited by 66
“Filling this development gap, the proposed approach in this article can bring significant results for monitoring air pollution by vehicles and even within smart cities scenarios, supporting a new development trend.”
BackgroundA Survey of the RISC-V Architecture Software Support
Mezger, B. W., Santos, D. A., Dilillo, L., Zeferino, C. A., & Melo, D. R. (2022) · IEEE Access · doi:10.1109/access.2022.3174125 · Cited by 86
“Furthermore, the work [77] provides a library to run lightweight and energy-efficient neural networks.”
BackgroundMethodIbba, P., Tronstad, C., Moscetti, R., Mimmo, T., Cantarella, G., Petti, L., Martinsen, Ø. G., Cesco, S., & Lugli, P. (2021) · Scientific Reports · doi:10.1038/s41598-021-90471-5 · Cited by 66
Wang, X., Hersche, M., Magno, M., & Benini, L. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3353146 · Cited by 72
“2) We reduce the input dimensions by automatically selecting the most relevant EEG channels to limit further the model complexity and the memory footprint while maintaining similar classification accuracy (82.79% versus 82.51% for the two-class task of the MM/MI dataset) or even improving the…”
MethodInfluentialSelf-citationPerformance Evaluation of Deep Learning Compilers for Edge Inference
Verma, G., Gupta, Y., Malik, A. M., & Chapman, B. (2021) · IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum · doi:10.1109/ipdpsw52791.2021.00128 · Cited by 80
“System on a Chip (SoC) [33] designs are another notable effort to improve the efficiency of inference on edge.”
BackgroundGómez-Carmona, O., Casado-Mansilla, D., Kraemer, F. A., López-de-Ipiña, D., & García-Zubia, J. (2020) · Future Generation Computer Systems · doi:10.1016/j.future.2020.06.013 · Cited by 52
“Moreover, Wang et al. [34] optimized the performance of multilayer perceptron meth-ods on low power embedded devices.”
Enabling Design Methodologies and Future Trends for Edge AI: Specialization and Codesign
Hao, C., Dotzel, J., Xiong, J., Benini, L., Zhang, Z., & Chen, D. (2021) · IEEE Design and Test · doi:10.1109/mdat.2021.3069952 · Cited by 67
Zhang, Y., Adın, V., Bader, S., & Oelmann, B. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3307751 · Cited by 56
“(TinyML) [29] have made it possible to bring deep learning”
BackgroundSchneider, T., Wang, X., Hersche, M., Cavigelli, L., & Benini, L. (2020) · IEEE International Conference on Smart Computing · doi:10.1109/smartcomp50058.2020.00065 · Cited by 58
“FANN-ON-MCU [11] is a different framework for exporting optimized neural networks to ARM processors, as well as to PULP-based systems.”
MethodSelf-citationLattanzi, E., Donati, M., & Freschi, V. (2022) · Sensors · doi:10.3390/s22072637 · Cited by 54
Moosmann, J., Giordano, M., Vogt, C., & Magno, M. (2023) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas57966.2023.10168657 · Cited by 80
“Thus, object detection networks need to be carefully designed and optimized for such tiny devices in particular to achieve a small memory footprint and a high number of operations processed per cycle [7].”
BackgroundSelf-citationA Battery-Free Long-Range Wireless Smart Camera for Face Detection
Giordano, M., Mayer, P., & Magno, M. (2020) · Proceedings of the 8th International Workshop on Energy Harvesting and Energy-Neutral Sensing Systems · doi:10.1145/3417308.3430273 · Cited by 38
Self-citationDNN is not all you need: Parallelizing Non-Neural ML Algorithms on Ultra-Low-Power IoT Processors
Tabanelli, E., Tagliavini, G., & Benini, L. (2021) · arXiv · doi:10.48550/arxiv.2107.09448 · Cited by 36
“[31] implemented an open-source framework to allow energy-efficient shallow ANNs execution on PULP processors.”
MethodA Directional Gamma-Ray Spectrometer With Microcontroller-Embedded Machine Learning
Buonanno, L., Vita, D. D., Carminati, M., & Fiorini, C. E. (2020) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2020.3029570 · Cited by 37
“microcontrollers have demonstrated to be very promising for energy-efficient ML at the Edge [17].”
BackgroundDNN Is Not All You Need: Parallelizing Non-neural ML Algorithms on Ultra-low-power IoT Processors
Tabanelli, E., Tagliavini, G., & Benini, L. (2023) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3571133 · Cited by 56
“[31] implemented an open-source framework to allow energy-efficient shallow ANNs execution on PULP processors.”
MethodAn evaluation methodology to determine the actual limitations of a TinyML-based solution
Delnevo, G., Mirri, S., Prandi, C., & Manzoni, P. (2023) · Internet of Things · doi:10.1016/j.iot.2023.100729 · Cited by 42
Wide-Band Label-Free Selective Microwave Resonator-Based Sensors for Multi-Component Liquid Analysis
Baghelani, M. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3137275 · Cited by 32
Survey and Comparison of Milliwatts Micro controllers for Tiny Machine Learning at the Edge
Giordano, M., Piccinelli, L., & Magno, M. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9870017 · Cited by 42
“Authors in [5] provide a comparison in terms of latency and power consumption of the two architectures, Cortex-M4 and PULP, over three different applications.”
BackgroundSelf-citationReal-Time Multiple Object Visual Tracking for Embedded GPU Systems
Fernández-Sanjurjo, M., Mucientes, M., & Brea, V. M. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3056239 · Cited by 41
“ASICs and FPGA devices are the best option for embedded computer vision with low footprint and power consumption [70]–[74].”
BackgroundChae, H. Y., Cho, J., Purbia, R., Park, C. S., Kim, H., Lee, Y., Baik, J. M., & Kim, J. J. (2022) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2022.3220871 · Cited by 32
“to include the concept of edge computing in sensor devices [21].”
BackgroundA Holistic Review of the TinyML Stack for Predictive Maintenance
Njor, E., Hasanpour, M. A., Madsen, J., & Fafoutis, X. (2024) · IEEE Access · doi:10.1109/access.2024.3512860 · Cited by 53
“This library takes models made in the Fast Artificial Neural Network (FANN) library and generates code that can run on ARM Cortex-M or Parallel Ultra Low Power (PULP) processors [64].”
CBin-NN: An Inference Engine for Binarized Neural Networks
Sakr, F., Berta, R., Doyle, J. C., Capello, A., Dabbous, A., Lazzaroni, L., & Bellotti, F. (2024) · Electronics · doi:10.3390/electronics13091624 · Cited by 27
“Currently, the most widespread edge devices are microcontrollers, which are typically characterized by energy efficiency and low costs [3].”
BackgroundMethodResearch progress on low-power artificial intelligence of things (AIoT) chip design
Ye, L., Wang, Z., Jia, T., Ma, Y., Shen, L., Zhang, Y., Li, H., Chen, P., Wu, M., Liu, Y., et al. (2023) · Science China Information Sciences · doi:10.1007/s11432-023-3813-8 · Cited by 30
“One common method for processing IoT data is to use low-power MCU chips, which offer good versatility and can handle various demands at the terminal side 4) [5].”
MethodFischer, R., Mueller, H., Polonelli, T., Benini, L., & Magno, M. (2021) · 2021 4th IEEE International Conference on Industrial Cyber-Physical Systems (ICPS) · doi:10.1109/icps49255.2021.9468256 · Cited by 21
“In the last decade, artificial intelligence has been considered a powerful data-driven technology to model highly complex problems and it has demonstrated its great potential to solve many data-driven problems [10].”
BackgroundSelf-citationRISC-HD: Lightweight RISC-V Processor for Efficient Hyperdimensional Computing Inference
Taheri, F., Bayat-Sarmadi, S., & Hadayeghparast, S. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3191717 · Cited by 30
“ble architecture of the RI5CY core, several works enhance this ISA for neural network and machine learning applications [27]–[30].”
BackgroundChen, J., Jiang, M., Zhang, X., da Silva, D. S., de Albuquerque, V. H. C., & Wu, W. (2022) · Expert Systems with Applications · doi:10.1016/j.eswa.2022.119407 · Cited by 24
Lalapura, V. S., Bhimavarapu, V. R., Amudha, J., & Satheesh, H. S. (2024) · Algorithms · doi:10.3390/a17030104 · Cited by 30
High-Speed Post-Quantum Cryptoprocessor Based on RISC-V Architecture for IoT
Hadayeghparast, S., Bayat-Sarmadi, S., & Ebrahimi, S. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3152850 · Cited by 19
“Moreover, the core has proven its practical efficiency through several implementations in recent work [23]–[25].”
BackgroundHersche, M., Rella, E. M., Mauro, A. D., Benini, L., & Rahimi, A. (2020) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3370748.3406560 · Cited by 25
“The MLP is mapped with the open-source toolkit FANN-on-MCU [23], that uses fixed-point arithmetic that also negligibly increases ARPE to 0.1575.”
MethodInfluentialAdil, M., Ali, A., Tin, T. T., Abulkasim, H., Farouk, A., Al-Kuwari, S. M., Song, H., & Jin, Z. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3605040 · Cited by 22
“Moreover, these devices operate on low-power micro-controllers (e.g., ARM Cortex-M0/M3, RISC-V cores) with limited CPU speeds (typically 100 MHz) and RAM ranging from 32 KB to 256 [184].”
Developing a TinyML Image Classifier in an Hour
Berta, R., Dabbous, A., Lazzaroni, L., Pau, D. P., & Bellotti, F. (2024) · IEEE Open Journal of the Industrial Electronics Society · doi:10.1109/ojies.2024.3451959 · Cited by 21
“Other significant examples, particularly for tiny devices, include FANN-on-MCU [14], edge learning machine [15], MicroAI [16], IoTSim-Edge [8], embedded learning library [17], edge impulse [18], and uTensor [19].”
BackgroundEnergy-Efficient Hardware-Accelerated Synchronization for Shared-L1-Memory Multiprocessor Clusters
Glaser, F., Tagliavini, G., Rossi, D., Haugou, G., Huang, Q., & Luca Benini, C. (2020) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2020.3028691 · Cited by 17
Sudharsan, B., Yadav, P., Breslin, J. G., & Ali, M. I. (2021) · IEEE International Conference on Services Computing · doi:10.1109/scc53864.2021.00045 · Cited by 23
“FANN-on-MCU [52] is an open-source framework that is built upon the Fast Artificial Neural Network (FANN) library [53] that can run light NNs on MCUs.”
MethodMachine Learning In-Sensors: Computation-enabled Intelligent Sensors For Next Generation of IoT
Ronco, A., Schulthess, L., Zehnder, D., & Magno, M. (2022) · Proceedings of IEEE Sensors/Proceedings of IEEE Sensors ... · doi:10.1109/sensors52175.2022.9967240 · Cited by 16
“Artificial Intelligence and Machine Learning paradigms are increasing their presence in the area of the Internet of Things (IoT) and Smart Sensors to face the challenging task of automatically extracting information from the data generated by ”traditional” sensors spread in the physical world [1].”
BackgroundSelf-citationCORDIC Hardware Acceleration Using DMA-Based ISA Extension
Manor, E., Ben-David, A., & Greenberg, S. (2022) · Journal of Low Power Electronics and Applications · doi:10.3390/jlpea12010004 · Cited by 16
“[5] presented efficient usage of custom instructions targeted at neural network processing.”
MethodWang, X., Geiger, F., Niculescu, V., Magno, M., & Benini, L. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3165828 · Cited by 25
Self-citationSun, T., Yue, W., Sun, T., Niu, H., & Li, Y. (2025) · Advanced Functional Materials · doi:10.1002/adfm.202503841 · Cited by 14
Lattanzi, E., Calisti, L., & Capellacci, P. (2023) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2023.101848 · Cited by 16
Polonelli, T., Schulthess, L., Mayer, P., Magno, M., & Benini, L. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401362 · Cited by 22
“…are available through two different tools: X-CUBE-AI from ST Microelectronics is an expansion package that extends the standard STM32 capabilities with automatic conversion of pre-trained Neural Network and ANSI C code generation, and FANN-ON-MCU, ARM Cortex-M, and low power micro-controllers [14].”
MethodSelf-citationMechanical Anomaly Detection on an Embedded Microcontroller
Lord, M., & Kaplan, A. (2021) · 2021 International Conference on Computational Science and Computational Intelligence (CSCI) · doi:10.1109/csci54926.2021.00159 · Cited by 13
“Wang et al presented an open-source toolkit built upon the fast artificial neural network (FANN) library to run lightweight and energy-efficient neural networks on MCUs based on both the ARM Cortex-M series and the novel RISC-V-based parallel ultralow-power platform [14].”
MethodExploring Automatic Gym Workouts Recognition Locally on Wearable Resource-Constrained Devices
Bian, S., Wang, X., Polonelli, T., & Magno, M. (2022) · International Green and Sustainable Computing Conference · doi:10.1109/igsc55832.2022.9969370 · Cited by 25
“GAP8 architecture, which features 8 RISC-V cores, has shown impressive inference performance in certain human activity recognition tasks [4], [25].”
BackgroundSelf-citationPham, N., Jia, H., Tran, M., Dinh, T., Bui, N., Kwon, Y., Ma, D., Nguyen, P., Mascolo, C., & Vu, T. (2022) · Proceedings of the 28th Annual International Conference on Mobile Computing And Networking · doi:10.1145/3495243.3560533 · Cited by 12
“MCUNet [45] mitigates the memory bottleneck of the CNN architecture with the patch-based inference approach,while FANN-MCU [100] provides toolkits for building energy-efficient networks on MCUs.”
BackgroundMethodOn Efficiency, Fairness and Security in AI Accelerator Resource Sharing: A Survey
Huang, J., Lin, W., Wu, W., Wang, Y., Zhong, H., Wang, X., & Li, K. (2025) · ACM Computing Surveys · doi:10.1145/3721427 · Cited by 19
“By leveraging specialized hardware [25], edge computing [59, 122], dynamic power management [78], and software optimization [30, 113], it is possible to significantly reduce the power consumption of tasks, ensuring efficient and sustainable AI deployments across various environments. where: —”
Self-citationLi, Z., Wang, F., Xie, H., Ke, D., Ye, T., Davari, S. A., & Kennel, R. M. (2023) · IEEE Transactions on Energy Conversion · doi:10.1109/tec.2023.3331580 · Cited by 13
“The first is by using tiny ML, which enables real-time diagnosis and protection of specific faults in micro-controller units (MCUs) with a power level of a few megawatts [34], [35].”
BackgroundHuang, B., Abtahi, A., & Aminifar, A. (2024) · IEEE transactions on circuits and systems for artificial intelligence. · doi:10.1109/tcasai.2024.3493036 · Cited by 20
“However, current resource-aware NAS only pays attention to individual hardware devices such as Microcontroller (MCU) boards [21], [22], [23], [24], [25], [26], mobile devices [16], [17], [27], server Central Processing Units (CPUs) and Graphics Processing Cards (GPUs) [16], [17], [20].”
InfluentialTCN Mapping Optimization for Ultra-Low Power Time-Series Edge Inference
Burrello, A., Dequino, A., Pagliari, D. J., Conti, F., Zanghieri, M., Macii, E., Luca Benini, C., & Poncino, M. (2021) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1109/islped52811.2021.9502494 · Cited by 12
“handling only fully-connected layers [13], or are closed source [14].”
BackgroundOnline Processing of Vehicular Data on the Edge Through an Unsupervised TinyML Regression Technique
de Andrade, P. H. M., da Silva, I. M. D., Diniz, M., Flores, T. K. S., Costa, D. G., & Soares, E. A. (2023) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3591356 · Cited by 16
“Quantized neural networks adapted to resource-constrained devices have also been implemented in other contexts, such as for detecting feeding events [51], keyword spotting applications [15, 18], for multiple purpose uses [74, 77] and IoT applications [10, 35, 46].”
MethodEvaluating Spiking Neural Network on Neuromorphic Platform For Human Activity Recognition
Bian, S., & Magno, M. (2023) · International Symposium on Wearable Computers · doi:10.1145/3594738.3611369 · Cited by 13
“With the increasing popularity of wearable devices, more complex HAR tasks have been moved from the laboratory to real-life scenarios powered by local AI solutions [3, 5, 36].”
BackgroundSelf-citationDesign of a 2-Bit Neural Network Quantizer for Laplacian Source
Perić, Z., Savić, M., Simić, N., Denić, B., & Despotović, V. (2021) · Entropy · doi:10.3390/e23080933 · Cited by 11
Compact and Stable Memristive Visual Geometry Group Neural Network
Ran, H., Wen, S., Li, Q., Cao, Y., Shi, K., & Huang, T. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3104860 · Cited by 12
“Edge computing has conducted in-depth research in academia and has been widely used in industry [9], [10].”
BackgroundLi, Y., Wu, Y., Zhang, X., Hu, J., & Lee, I. (2022) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2022.3171308 · Cited by 17
Rutishauser, G., Hunziker, R., Mauro, A. D., Bian, S., Benini, L., & Magno, M. (2023) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas46773.2023.10181726 · Cited by 16
“Kraken includes a heterogeneous parallel ultra-low power (PULP) processor [18] and two hardware accelerators for both spiking neural networks and frame-based convolutional neural networks.”
MethodSelf-citationPapa, L., Mattia, G. P., Russo, P., Amerini, I., & Beraldi, R. (2023) · Sensors · doi:10.3390/s23042223 · Cited by 11
“Instead, in [23], the authors propose an energy-efficient toolkit for performing inference in Edge Computing.”
BackgroundSmartHand: Towards Embedded Smart Hands for Prosthetic and Robotic Applications
Wang, X., Geiger, F., Niculescu, V., Magno, M., & Luca Benini, C. (2021) · IEEE Sensors Applications Symposium (SAS) · doi:10.1109/sas51076.2021.9530050 · Cited by 10
Self-citationAn efficient design methodology to speed up the FPGA implementation of artificial neural networks
Vineetha, K. V., Reddy, M. M. S. K., Ramesh, C., & Kurup, D. G. (2023) · Engineering Science and Technology an International Journal · doi:10.1016/j.jestch.2023.101542 · Cited by 10
Wiedemann, S., Shivapakash, S., Becking, D., Wiedemann, P., Samek, W., Gerfers, F., & Wiegand, T. (2021) · IEEE Open Journal of Circuits and Systems · doi:10.1109/ojcas.2021.3083332 · Cited by 13
“Moreover, multilayer perceptrons (MLPs) are already the status quo in use cases with very tight resource constraints, since many studies identified MLPs to be one of the best algorithms to solve tasks in the IoT domain using wearable devices [2].”
BackgroundSelective Neuron Re-Computation (SNRC) for Error-Tolerant Neural Networks
Liu, S., Reviriego, P., & Lombardi, F. (2021) · IEEE Transactions on Computers · doi:10.1109/tc.2021.3056992 · Cited by 11
Vitelli, M., Cerro, G., Gerevini, L., Miele, G., Ria, A., & Molinara, M. (2023) · Computers · doi:10.3390/computers12020023 · Cited by 11
Enabling ImageNet-Scale Deep Learning on MCUs for Accurate and Efficient Inference
Sadiq, S., Hare, J. S., Craske, S., Maji, P., & Merrett, G. V. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3331654 · Cited by 12
“The frameworks use optimized kernels to efficiently execute operations in the inference graph [13], [15], [37], [38].”
BackgroundMethodLi, Q., Li, H., & Meng, L. (2024) · Journal of Real-Time Image Processing · doi:10.1007/s11554-024-01496-8 · Cited by 11
“However, feedback from AIoT systems usually has unacceptable latency due to the limited bandwidth of the network and instability of communication [2, 3].”
EdgeMark: An automation and benchmarking system for embedded artificial intelligence tools
Hasanpour, M. A., Kirkegaard, M., & Fafoutis, X. (2025) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2025.103488 · Cited by 17
An 8-bit Single Perceptron Processing Unit for Tiny Machine Learning Applications
Crepaldi, M., Salvo, M. D., & Merello, A. (2023) · IEEE Access · doi:10.1109/access.2023.3327517 · Cited by 10
“To provide a more thorough comparison with the RISC-V-based platform presented in [30], we have considered 119916 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.”
BackgroundMethodInfluentialGiordano, M., Fischer, R., Crabolu, M., Bellusci, G., & Magno, M. (2021) · IEEE Sensors Applications Symposium (SAS) · doi:10.1109/sas51076.2021.9530182 · Cited by 11
Self-citationRPR: Random Partition Relaxation for Training; Binary and Ternary Weight Neural NetworksPreprint
Cavigelli, L., & Benini, L. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2001.01091 · Cited by 12
“At the same time, recent research into specialized hardware accelerators has shown that improvements by 10–100× in energy efficiency over optimized software are achievable [19]– [22].”
BackgroundSelf-citationAn Energy-Efficient Near-Memory Computing Architecture for CNN Inference at Cache Level
Nouripayam, M., Prieto, A., Kishorelal, V. K., & Rodrigues, J. N. (2021) · IEEE International Conference on Electronics, Circuits and Systems (ICECS) · doi:10.1109/icecs53924.2021.9665530 · Cited by 12
Dynamic ConvNets on Tiny Devices via Nested Sparsity
Grimaldi, M., Mocerino, L., Cipolletta, A., & Calimera, A. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3222014 · Cited by 13
“putational graph rewritings [8], [9], [10]; and computational kernel tuning [9], [11], [12].”
BackgroundTiny-FPU: Low-Cost Floating-Point Support for Small RISC-V MCU Cores
Bertaccini, L., Perotti, M., Mach, S., Schiavone, P. D., Zaruba, F., & Luca Benini, C. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401149 · Cited by 7
“We used a synthetic benchmark (synth), a patternrecognition algorithm knn, a 2D convolution conv2d, a ReLU kernel, an axpy algorithm, and an excerpt of the FANN library [12] in two versions: integer-only (fann-int) and FP-only (fannfp).”
MethodEnergy-aware Tiny Machine Learning for Sensor-based Hand-washing Recognition
Lattanzi, E., & Calisti, L. (2023) · International Conference on Machine Learning Technologies (ICMLT) · doi:10.1145/3589883.3589886 · Cited by 7
“proposed a tool to design an energy-efficient multilayer Perceptron tailored for 1MB of RAM microcontrollers [45] while”
MethodWang, X., Schneider, T., Hersche, M., Cavigelli, L., & Benini, L. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401564 · Cited by 10
“body of the user, using low-power low-cost microcontroller units (MCUs), allowing the data to be processed locally [16].”
BackgroundSelf-citationBringing In-Sensor Intelligence in Radiation Detectors: a Short Review
Carminati, M., Giacomo, S. D., Ronchi, M., Borghi, G., & Fiorini, C. E. (2023) · IEEE EUROCON - International Conference on Computer as a Tool (EUROCON) · doi:10.1109/eurocon56442.2023.10198991 · Cited by 6
“The most common approach in this field is to migrate processing from PC to embedded digital devices such as FPGA and MCU, Fig.”
MethodInfluentialEfficient Transform Algorithms for Parallel Ultra-Low-Power IoT End Nodes
Mazzoni, B., Benatti, S., Luca Benini, C., & Tagliavini, G. (2021) · IEEE Embedded Systems Letters · doi:10.1109/les.2021.3065206 · Cited by 6
A data-stream TinyML compression algorithm for vehicular applications: a case study
Silva, M., Signoretti, G., Flores, T. K. S., de Andrade, P. H. M., Silva, J., da Silva, I. M. D., Sisinni, E., & Ferrari, P. (2022) · IEEE International Workshop on Metrology for Industry 4.0 & IoT (MetroInd4.0&IoT) · doi:10.1109/metroind4.0iot54413.2022.9831606 · Cited by 6
Dincic, M. R., Perić, Z., Denić, D. B., & Denić, B. (2023) · Measurement · doi:10.1016/j.measurement.2023.113037 · Cited by 7
A Review of Edge Intelligence Applications Based on RISC-V
Qian, W., Cui, E., Gao, Y., & Li, T. (2023) · International Conference on Computing, Communication, Perception and Quantum Technology (CCPQT) · doi:10.1109/ccpqt60491.2023.00025 · Cited by 6
Configurable Hardware Acceleration for Hyperdimensional Computing Extension on RISC-V
Martino, R., Angioli, M., Rosato, A., Barbirotta, M., Cheikh, A., & Olivieri, M. (2025) · IEEE Transactions on Computers · doi:10.1109/tc.2025.3642250 · Cited by 5
“Several works in the literature have designed custom RISC-V extensions for neural networks and machine learning applications [30], [31].”
Bosse, S. (2023) · Annals of Computer Science and Information Systems · doi:10.15439/2023f7745 · Cited by 6
Giordano, M., Cortesi, S., Crabolu, M., Pedrollo, L., Bellusci, G., Bendinelli, T., Türetken, E., Dunbar, A. L., & Magno, M. (2023) · IEEE World Forum on Internet of Things (WF-IoT) · doi:10.1109/wf-iot58464.2023.10539550 · Cited by 8
“The advancements in machine learning (ML) have enabled the development of solutions that offer accurate detection using a variety of algorithms that operate on sensor data [11], [12].”
MethodSelf-citationSub-100 $\mu$W Multispectral Riemannian Classification for EEG-Based Brain–Machine Interfaces
Wang, X., Cavigelli, L., Schneider, T., & Benini, L. (2021) · IEEE Transactions on Biomedical Circuits and Systems · doi:10.1109/tbcas.2021.3137290 · Cited by 10
Self-citationPicoSAM2: Low-Latency Segmentation In-Sensor for Edge Vision Applications
Bonazzi, P., Farronato, N., Zihlmann, S., Qin, H., & Magno, M. (2025) · Proceedings of IEEE Sensors/Proceedings of IEEE Sensors ... · doi:10.1109/sensors59705.2025.11330465 · Cited by 4
Self-citation
Metadata
- Authors
- Wang, X., Magno, M., Cavigelli, L., & Benini, L.
- Affiliations
- ETH Zurich (CH)
- Venue
- IEEE Internet of Things Journal
- Published
- 2020-05
- Volume
- 7
- Issue
- 5
- Pages
- 4403-4417
- ISSN
- 2327-4662 (ISSN-L)
- 2372-2541
- OpenAlex
- W3007729836
- Cited by
- 162
- References
- 59
- Citations by year
- 2026: 12 · 2025: 18 · 2024: 18 · 2023: 32 · 2022: 33 · 2021: 34 · 2020: 15
- FWCI
- 13.39
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Internet of Things Journal (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Archivio istituzionale della ricerca (Alma Mater Studiorum Università di Bologna) (submittedVersion, other-oa)
- arXiv (Cornell University)
- Funders
- Swiss Data Science Center Ph.D. Fellowship (ID P18-04)
- U.S. Office of Naval Research Global “Zero-Power Sensing for Underwater Monitoring,” (ONRG - NICOP - N62909-19-1-2018)
- Armasuisse Science and Technology
- Topics
- 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
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- edge computing, Internet of Things, microcontroller, ARM Cortex-M, RISC-V, multilayer perceptron, code generation, parallelization overhead, memory transfer, real-time classification
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.