The DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index; 28 of the 56 sources its claims cite were checked for support.
TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications
Open accessAbstract
Recently, the Internet of Things (IoT) has gained a lot of attention, since IoT devices are placed in various fields. Many of these devices are based on machine learning (ML) models, which render them intelligent and able to make decisions. IoT devices typically have limited resources, which restricts the execution of complex ML models such as deep learning (DL) on them. In addition, connecting IoT devices to the cloud to transfer raw data and perform processing causes delayed system responses, exposes private data and increases communication costs. Therefore, to tackle these issues, there is a new technology called Tiny Machine Learning (TinyML), that has paved the way to meet the challenges of IoT devices. This technology allows processing of the data locally on the device without the need to send it to the cloud. In addition, TinyML permits the inference of ML models, concerning DL models on the device as a Microcontroller that has limited resources. The aim of this paper is to provide an overview of the revolution of TinyML and a review of tinyML studies, wherein the main contribution is to provide an analysis of the type of ML models used in tinyML studies; it also presents the details of datasets and the types and characteristics of the devices with an aim to clarify the state of the art and envision development requirements.
Overview
CoChat Summary
TL;DR
This review surveys TinyML, the deployment of compressed deep-learning inference on ultra-low-power IoT edge devices such as microcontrollers, and analyzes models, datasets, devices, frameworks, and limitations across 38 selected papers. It finds that TinyML offers energy, cost, latency, and privacy benefits, but remains constrained by device heterogeneity, limited processing power and memory, and a shortage of suitable DL models.
Key findings
- Across the reviewed TinyML studies, image datasets were the most common data type at 46%, followed by physiological/behavioral metrics at 29% and audio at 11%; the most frequently used image datasets were ImageNET and Visual Wake Words, while Google Speech Commands was the most used audio dataset.
- The most common models achieving best results were ANN at 21%, CNN at 12%, and MobileNet-V1 at 6%; non-neural ML models included Decision Tree at 9%, SVM at 6%, and KNN at 6%.
- Device usage was dominated by STM32 microcontrollers: STM32-F746ZG appeared in 22% of studies and STM32-H743ZI2 in 20%; reviewed device costs ranged approximately from USD 16.00 to USD 28.00, and MCUs were described as typically small (~1 cm³), low power (~1 mW), and cheap (~$1).
- Reported use-case results included ASL detection at 98.84% accuracy and 20 FPS after deployment, handwriting recognition at 100% prediction accuracy with 135.68 KB RAM, 668.97 KB flash, and 330 ms computation, medical face mask detection at 99.83% accuracy and 30 FPS with a 138 KB model, and MCUNet reaching 70.0% accuracy on STM32H743 with 0.49 MB SRAM and 0.9 MB flash.
Methods
- This is a systematic literature review following PRISMA guidelines, organized around research questions on datasets, models, devices, application domains, frameworks/libraries, and limitations in TinyML.
- Searches used Google Scholar, Clarivate’s Web of Science, Science Direct, IEEE, MDPI, and Springer; the first TinyML publication was dated 13 December 2019, and a total of 38 relevant published papers/articles were retrieved and screened using inclusion/exclusion criteria.
- Data extraction covered each study’s main contribution, implementation details including datasets, models, devices, and evaluation criteria; findings were synthesized descriptively into dataset, model, and device categories.
Limitations
- The paper states that TinyML faces device heterogeneity, with no standardized framework across manufacturers; lower processing power and speed than cloud-based systems; limited memory of only a few KB; and a lack of DL models that achieve high accuracy with small model size and low latency.
- Evident from the text: as a review, the findings are descriptive counts and summaries from 38 selected studies rather than new experimental validation, and the exclusion criteria omitted studies on compression techniques such as quantization and pruning, memory reduction, TinyML services, technical reports, and website-published experiments.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures






Verification
Claims the paper makes
23 Valid · 9 Not valid · 19 Not checked · 35 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, edge devices have not yet become cost-efficient enough to take training DL models due to limited resources [7,8,9].”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 1. Introduction
What the sources say
- Not checked — no full text of the sourcecites [7] Atitallah, 2020, Ben Leveraging Deep Learning and IoT big data analytics to support the smart cities development: Review and future directions, Comput. Sci. Rev. source
- Not checked — no full text of the sourcecites [8] Bhattacharya, 2020, A review on deep learning for future smart cities, Internet Technol. Lett. source
- Partly supportedcites [9] Wang, 2020, Deep learning for edge computing applications: A State-of-the-Art survey, IEEE Access sourceExperimental — accuracy not yet measured
The Model Training section says training deep models consumes massive CPU and GPU resources and that edge servers are usually challenged, or not cost-efficient, when undertaking training alone. The claim broadens this from edge servers to edge devices generally and omits the qualification that the servers undertake training solely; the paper distinguishes devices from edge servers in Model Inference. Page numbers are unavailable in our copy. (LLM, conf=0.86)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Authors in [28] proposed a model to detect the American Sign language (ASL) Alphabet and transcribed to text and speech in real time on tiny wearable IoT devices.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 4. Related Work to TinyML > 4.1. TinyML Use Cases Studies
What the sources say
- Partly supportedcites [28] Paul, A.J., Mohan, P., and Sehgal, S. (2020, January 3–5). Rethinking Generalization in American Sign Language Prediction for Edge Devices with Extremely Low Memory Footprint. Proceedings of the 2020 IEEE Recent Advances in Intelligent Computational Systems, RAICS, Thiruvananthapuram, India. sourceExperimental — accuracy not yet measured
The Introduction (p. 1) describes a proposed ASL fingerspelling model small enough for wearable-IoT microcontrollers and envisions real-time transcription to text or speech. The Conclusion (p. 6) confirms real-time recognition on an OpenMV H7 board, but says a watch-like wearable device remains future work; our copy does not establish that text-and-speech transcription was implemented. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“It is computationally expensive, consuming massive CPU and GPU resources, power, memory, and time.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 1. Introduction
What the sources say
- Not checked — no full text of the sourcecites [7] Atitallah, 2020, Ben Leveraging Deep Learning and IoT big data analytics to support the smart cities development: Review and future directions, Comput. Sci. Rev. source
- Not checked — no full text of the sourcecites [8] Bhattacharya, 2020, A review on deep learning for future smart cities, Internet Technol. Lett. source
- Partly supportedcites [9] Wang, 2020, Deep learning for edge computing applications: A State-of-the-Art survey, IEEE Access sourceExperimental — accuracy not yet measured
The Model Training section says training is often computation-intensive and consumes massive CPU and GPU resources. It does not specifically attribute substantial power, memory, or time consumption to training; the passage about time and energy concerns model inference instead. Page numbers are unavailable in our copy. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Furthermore, deployment of pre-trained DL models into tiny edge devices is now possible, after performing some techniques to compress DL models and optimize the inference.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 1. Introduction
What the sources say
- Partly supportedcites [15] Wang, 2020, Pruning from Scratch, Proceedings of the AAAI Conference on Artificial Intelligence sourceExperimental — accuracy not yet measured
The Introduction (p. 1) says deep neural networks are deployed on mobile devices and that pruning compresses models and accelerates inference. It does not establish that these techniques make deployment of pre-trained models on tiny edge devices possible; the paper instead emphasizes pruning without pre-training. (LLM, conf=0.87)
- Supportedcites [16] Wardana, I.N.K., Gardner, J.W., and Fahmy, S.A. (2021). Optimising Deep Learning at the Edge for Accurate Hourly Air Quality Prediction. Sensors, 21. source
The Introduction describes compressing trained models for edge deployment and identifies post-training quantisation as a way to reduce model size and potentially improve latency. Section 3.5 reports that the authors optimised a trained model and deployed it to Raspberry Pi edge devices; Section 5.3 reports reductions in model size. (LLM, conf=0.90)
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, inference of DL in edge devices is not yet prepared enough to be fully executed [3,4,6].”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 1. Introduction
What the sources say
- Partly supportedcites [3] Signoretti, G., Silva, M., Andrade, P., Silva, I., Sisinni, E., and Ferrari, P. (2021). An Evolving Tinyml Compression Algorithm for IoT Environments Based on Data Eccentricity. Sensors, 21. sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) says that computational restrictions lead TinyML researchers to reduce and optimize neural-network models for tiny devices. That supports a limitation on edge-device inference, but not the broader assertion that deep-learning inference is not yet ready to be fully executed on edge devices; the same section describes ML models being deployed there. (LLM, conf=0.84)
- Not checked — no full text of the sourcecites [4] Chen, 2020, Deep Learning on Mobile and Embedded Devices: State-of-the-Art, Challenges, and Future Directions, ACM Comput. Surv. source
- Supportedcites [6] Gorospe, J., Mulero, R., Arbelaitz, O., Muguerza, J., and Antón, M.Á. (2021). A Generalization Performance Study Using Deep Learning Networks in Embedded Systems. Sensors, 21. source
The Introduction states that DNNs are not prepared to be fully executed in edge-computing devices because their computing and memory demands exceed those devices’ resources. This directly supports the cited claim about on-device deep-learning inference. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
18 more claims · 1 with their cited sources checked
Create a free account to see all 86 claims and to verify the 8 references still open.
- Probably not a substantial claim. 45
- It may state two separate claims as one. 39
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 3. Research Methodology > 3.2. Search Strategy
“The first publication for TinyML was on 13 December 2019.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 3. Research Methodology > 3.6. Data Selection
“The first published paper on the emerging topic was on 19 December 2019”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 4. Related Work to TinyML > 4.1. TinyML Use Cases Studies
“the best performance was that of ANN in all datasets except Sonar dataset.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 4. Related Work to TinyML > 4.1. TinyML Use Cases Studies
“In contrast the SVM achieved the best results on the Heart and EnviroCar dataset.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 3. Research Methodology > 3.5. Data Extraction and Synthesis
“The information extracted from studies is described in Section 6.”
§ TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications > 5. Discussion and Findings
“## 5. Discussion and Findings Based on the related works of tinyML, we have derived the following findings concerning models, datasets, and devices.”
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
The PRISMA 2020 Statement: An Updated Guideline for Reporting Systematic Reviews
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., et al. (2021) · International Journal of Surgery · doi:10.1016/j.ijsu.2021.105906 · Cited by 11,748
Automated Flower Classification over a Large Number of Classes
Nilsback, M.-E., & Zisserman, A. (2008) · 2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing · doi:10.1109/icvgip.2008.47 · Cited by 3,442
ImageNet: A large-scale hierarchical image database
Deng, J. H., Dong, W. F., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvprw.2009.5206848 · Cited by 2,714
Parkhi, O. M., Vedaldi, A., Zisserman, A., & Jawahar, C. V. (2012) · 2012 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2012.6248092 · Cited by 1,533
Analysis of hidden units in a layered network trained to classify sonar targets
Gorman, R. P., & Sejnowski, T. J. (1988) · Neural Networks · doi:10.1016/0893-6080(88)90023-8 · Cited by 1,114 · cited 1×
“Data Virus dataset [54] Sonar dataset [55,56] Peugeot 14 [57] Peugeot 15 [57] EnviroCar [58] Air Quality Index (AQI) [59] Dset-2.”
MethodMachine Learning for Internet of Things Data Analysis: A Survey
Mahdavinejad, M. S., Rezvan, M., Barekatain, M., Adibi, P., Barnaghi, P. M., & Sheth, A. (2017) · Surrey Research Insight Open Access (The University of Surrey) · Cited by 1,008
Machine learning for internet of things data analysis: a survey
Mahdavinejad, M. S., Rezvan, M., Barekatain, M., Adibi, P., Barnaghi, P., & Sheth, A. P. (2018) · Digital Communications and Networks · doi:10.1016/j.dcan.2017.10.002 · Cited by 921 · cited 1×
“Developing a generalized framework is necessary for increasing adoption and awareness of TinyML [62,63].”
BackgroundIoT in smart cities: A survey of technologies, practices and challenges
Syed, A. S., Sierra-Sosa, D., Kumar, A., & Elmaghraby, A. (2021) · Smart Cities · doi:10.3390/smartcities4020024 · Cited by 527 · cited 1×
“The proliferation of IoT edge devices lead to collection of sensitive data from society, wherein transmission of data to the cloud may expose it to hacking and eavesdropping [13].”
BackgroundEdge Machine Learning for AI-Enabled IoT Devices: A Review
Merenda, M., Porcaro, C., & Iero, D. (2020) · Sensors · doi:10.3390/s20092533 · Cited by 523 · cited 4×
“[25] Handwritten Digits Handwritten digit images from (0 to 9) 70,000 images 60,000 10,000”
BackgroundMethodInfluential
Research works printed but not linked: 6
- Banbury, C.R., Reddi, V.J., Lam, M., Fu, W., Fazel, A., Holleman, J., Huang, X., Hurtado, R., Kanter, D., and Lokhmotov, A. (2020). Benchmarking TinyML Systems: Challenges and Direction. arXiv.
- Venzke, M., Klisch, D., Kubik, P., Ali, A., Missier, J.D., and Turau, V. (2020). Artificial Neural Networks for Sensor Data Classification on Small Embedded Systems. arXiv.
- Wong, A., Famouri, M., Pavlova, M., and Surana, S. (2020). TinySpeech: Attention Condensers for Deep Speech Recognition Neural Networks on Edge Devices. arXiv.
- Cai, H., Gan, C., Zhu, L., and Han, S. (2020). TinyTL: Reduce Memory, Not Parameters for Efficient On-Device Learning. arXiv.
- Vu, T.H., Olsson, C., Laptev, I., Oliva, A., and Sivic, J. (2014). Food-101–Mining Discriminative Components with Random Forests, Springer International Publishing. LNCS 8693.
- Chowdhery, A., Warden, P., Shlens, J., Howard, A., and Rhodes, R. (2019). Visual Wake Words Dataset. arXiv.
Cited by
Items 101–200 of 220
Zhang, S.-Z., Yang, H.-Z., & Gao, Y. (2026) · Scientific Reports · doi:10.1038/s41598-026-41195-x · Cited by 4
Ramkumar, G. (2026) · International Conference on Electronics and Renewable Systems (ICEARS) · doi:10.1109/icears67481.2026.11416596 · Cited by 3
“In [9], the author defines TinyML as technology enabling Machine Learning and Deep Learning on IoT devices with low resources, enhancing latency, privacy, and reducing communication costs.”
Moheth, A., Mulcahy, A., Akther, M., Akter, S., Azimi, B., Puccio, F. D., Danti, S., Alam, M. M., & Uddin, M. J. (2026) · Nano Trends · doi:10.1016/j.nwnano.2026.100199 · Cited by 3
Artificial intelligence enabled smart education systems
Sharma, A. K., & Kumar, T. V. V. (2026) · International Journal of Systems Assurance Engineering and Management · doi:10.1007/s13198-026-03196-x · Cited by 3
Multi-Objective Evaluation of Lightweight AI Models on Low-Cost Edge Devices Using Pareto Fronts
Rojas-Carrasco, P., & Guinaldo, M. (2026) · Applied Sciences · doi:10.3390/app16083679 · Cited by 2
Optimization and Benchmarking of Lightweight Neural Networks for Efficient Embedded AI Deployment
Fridous, V. M., Agarwal, A., Bhaskar, K. B., Nandagopal, V., & Sivakamasundari, N. (2026) · Engineering Reports · doi:10.1002/eng2.70814 · Cited by 5
Tiny Machine Learning Virtualization for IoT and Edge Computing using the REXA VM
Bosse, S., & Polle, C. (2023) · International Conference on Future Internet of Things and Cloud (FiCloud) · doi:10.1109/ficloud58648.2023.00026 · Cited by 2
“Ultra low-power devices introduce additional constraints on the computation of Deep learning (DL) models [5], and hardware designs gains attraction [6], which are addressed in this work, too.”
BackgroundCompiler-Based Memory Encryption for Machine Learning on Commodity Low-Power Devices
Maeng, K., & Lucia, B. (2024) · ACM SIGPLAN International Conference on Compiler Construction (CC) · doi:10.1145/3640537.3641564 · Cited by 3
LogicNets vs. ULEEN : Comparing two novel high throughput edge ML inference techniques on FPGA
Nag, S., Susskind, Z., Arora, A., Bacellar, A. T. L., Dutra, D. L. C., Miranda, I. D. S., Kailas, K. K., John, E. B., Breternitz, M., Lima, P. M. V., et al. (2024) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas60917.2024.10658913 · Cited by 2
“This has necessitated the need for high-throughput inferences of ML models at the edge [2].”
BackgroundDetecting Gesture Language for Deaf and Mute People Usingon Ultra-Low-Power TinyML Model
Dabwan, B. A., Jadhav, M. E., Ismil, O. A., Hassan, E. A., Farah, E. A., Mohammad, A. A., & Ali, Y. A. (2024) · International Conference on Data Science and Network Security (ICDSNS) · doi:10.1109/icdsns62112.2024.10691053 · Cited by 5
Dehnaw, A. M., Lu, Y.-J., Shih, J.-H., Yao, C.-K., Bitew, M. A., & Peng, P. (2024) · Processes · doi:10.3390/pr12122638 · Cited by 6
Toward Efficient Federated Load Forecasting: Personalization Mechanisms and Their Impact
Dissem, M., & Amayri, M. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3580378 · Cited by 5
“Consequently, the models can be efficiently supported by a wide range of inference devices [59], [60].”
AI-Assisted IoT Monitoring for Tracking Construction and Demolition Waste
Trávníček, P., & Nežerka, V. (2025) · International Conference on Smart and Sustainable Technologies (SpliTech) · doi:10.23919/splitech65624.2025.11091388 · Cited by 3
“Historically, such deep learning tasks were infeasible on resource-constrained devices such as the ESP32; however, advances in edge computing and tools like TinyML now enable these models to run efficiently at the edge [8]–[10].”
Smart Presence Detection: Harnessing Wi-Fi Signals and Machine Learning with ESP8266Preprint
Khan, R., & Khan, D. (2024) · Preprints.org · doi:10.20944/preprints202406.1346.v1 · Cited by 2
Smaller Together: Groupwise Encoding of Sparse Neural Networks
Trommer, E., Waschneck, B., & Kumar, A. S. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3402957 · Cited by 1
“Digital Object Identifier 10.1109/TCAD.2024.3402957 and HPC systems offer an abundance of memory at minimal latency through hierarchical layers of caches, MCUs are only equipped with small amounts of Flash and static random-access memory (SRAM) memory, usually in the range of KiB to a few MiB [3].”
InfluentialTinyML: principles and algorithms
Ghorpade, S. N., Chougule, S., Chaudhari, B. S., & Zennaro, M. (2024) · Elsevier eBooks · doi:10.1016/b978-0-44-322202-3.00009-9 · Cited by 3
Tiny Machine Learning for IoT-Enabled Embedded Systems: A Review
Mrabet, A. E., Tber, A., Elmousaid, R., Hlou, L., & gouri, R. E. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-93448-3_6 · Cited by 1
Mishra, R. K., Alansari, K., Cole, R., et al. (2025) · Sensors · doi:10.3390/s25154767 · Cited by 1
Human Activity Recognition on Embedded Devices: CNN-Based Walking Detection and Step Counting
Pascale, R., Tudose, D., & Alexandru, N. (2025) · Computer Science in Cars Symposium · doi:10.1109/cscs66924.2025.00058 · Cited by 1
Ali, N. F., Al-Nashash, H. A., Elfadel, I. M., & Atef, M. (2025) · IEEE Access · doi:10.1109/access.2025.3639253 · Cited by 2
“FPGA consumes less power and demonstrates acceptable performance for ML, but it is less efficient than ASIC [19], thus requiring a higher development cost [18].”
Rauniyar, A. K., Gupta, A., & Anitha, G. (2025) · International Conference on Software, Systems and Information Technology (SSITCON) · doi:10.1109/ssitcon66133.2025.11342102 · Cited by 1
“More recently, Alajlan and Ibrahim [22] reviewed TinyML strategies, showing that model pruning and quantization are critical for making digit recognition feasible on ultra-low-power IoT edge devices.”
Yarham, S., Behjati, M., & Nordin, R. (2025) · International Conference on Computer, Internet of Things and Smart City (CIoTSC) · doi:10.1109/ciotsc67482.2025.11413126 · Cited by 1
“The most fundamental challenge is resource limitation, as microcontrollers often provide only a few kilobytes of RAM and less than a megabyte of flash memory, restricting the complexity of deployable models and frequently reducing accuracy compared to cloud-based systems [6], [8], [28], [29].”
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
PERFECG-Net: A Lightweight Periodicity-Focused Fetal ECG R-Peak Detection Network
Ameta, D., Somasundaram, S., Srivarshini, S., et al. (2026) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2026.3682832 · Cited by 1
“Deep learning models with millions of parameters are typically unsuitable for embedded platforms due to limitations in memory, latency, and energy consumption [21].”
Muscle myography for human–machine interfaces: a review of sensing modalities and control interfaces
Zhuwawu, S. S., Owen, M., Bifet, A., & Dwivedi, A. (2026) · Frontiers in Bioengineering and Biotechnology · doi:10.3389/fbioe.2026.1759542 · Cited by 3
“Frontiers in Bioengineering and Biotechnology frontiersin.org comes at the cost of reduced accuracy (Alajlan and Ibrahim, 2022).”
Rao, A. R., Prasad, C. R., & Mateti, J. (2026) · doi:10.1109/icsft67370.2026.11566295 · Cited by 1
“Both latency and reliability are crucial for audio applications [2], [3].”
Edge Intelligence in the IoT Era: A Review of Architectural Paradigms
Fiore, M., & Lanera, F. (2026) · Electronics · doi:10.3390/electronics15163689 · Cited by 1
Alharthi, R. S., & Aurangzeb, K. (2023) · International Conference on Bio-engineering for Smart Technologies (BioSMART) · doi:10.1109/biosmart58455.2023.10162079 · Cited by 1
“[10].”
Enhancement in IoT through Custom Instruction Set Architectures and TinyML: Review
Sandhya, P., & Chandran, P. (2023) · International Conference on Computer, Electronics & Electrical Engineering & their Applications (IC2E3) · doi:10.1109/ic2e357697.2023.10262472 · Cited by 1
“With TinyML, IoT devices can offer precise machine learning services while operating reliably without the need for continuous cloud service access [6].”
BackgroundRobust MCU Oriented KWS Model for Children Robotic Prosthetic Hand Control
Perdomo-Campos, A., Beltrán, J. R., & Morgado‐Estévez, A. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-49552-6_25 · Cited by 1
Jannesar, N., Akbarzadeh-Sherbaf, K., Safari, S., & Vahabie, A. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106368 · Cited by 3
Real-Time Speed Breaker Detection with an Edge Impulse
Manjula, S., Sudha, M. S., Yogaraja, C. A., et al. (2024) · SN Computer Science · doi:10.1007/s42979-024-03132-5 · Cited by 1
“The new technology called Tiny Machine Learning (TinyML), that has paved the way to meet the challenges of IoT devices [24].”
Deshmukh, S., Bende, A., Sanghai, D., Saraswat, V., Biswas, A., Kadam, A. A., Patil, S. V., Singh, A. K., Deshpande, V. V., & Ganguly, U. (2024) · IEEE Journal of the Electron Devices Society · doi:10.1109/jeds.2024.3508759 · Cited by 1
“This transition enables edge devices to operate independently of the cloud, facilitating faster response times for latency-sensitive applications [1], [5], [6].”
Evaluating Performance Trade-offs of Caching Strategies for AI-Powered Querying Systems
Oh, H.-J., Zhang, W., Rickett, C. D., Sukumar, S. R., & Byna, S. (2024) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata62323.2024.10825819 · Cited by 1
“The AI-inference process is known to require significant computing power, and generally it is a time-consuming process iterating over multiple phases [10], [32].”
Sub-Nyquist Harmonic Current Component Extraction using Band Pass Filters for NILM
Jung, G., Kim, G. Y., & Ahn, S. (2025) · International Journal of Precision Engineering and Manufacturing-Green Technology · doi:10.1007/s40684-024-00691-z · Cited by 3
“An SVM requires fewer computations and less memory than those required by other methods and has a fast inference time, making the SVM suitable for microcontrollers and real-time applications [34].”
Perspective Chapter: The Evolution of Edge Computing in the IoT Era of the 22nd Century
Nwokoma, F., & Akujuobi, C. M. (2024) · IntechOpen eBooks · doi:10.5772/intechopen.1006375 · Cited by 1
ELPG: End-to-End Latency Prediction for Deep-Learning Model Inference on Edge Devices
Zhao, W., Wang, L., & Duan, F. (2024) · International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc62609.2024.10942010 · Cited by 1
“Deep learning models have been widely used in image classification [1-2], object detection [3], and image segmentation [4]. the demand for deploying models to edge devices is increasing [5].”
Editorial for the Special Issue on Micro and Smart Devices and Systems
Mao, Z., Xie, J., & Ding, H. (2023) · Micromachines · doi:10.3390/mi14010164 · Cited by 1
“[21] focus on the Internet of Things (IoT) devices and use machine learning (ML) models to make intelligent decisions.”
BackgroundAutomated Speaker Independent Visual Speech Recognition: A Comprehensive SurveyPreprint
Nemani, P., Krishna, G. S., & Kundrapu, S. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2306.08314
“Edge device deployment [175] and real-time testing are important aspects of VSR systems.”
MethodIntelligent Shipping Using TinyML
(2023) · International Research Journal of Modernization in Engineering Technology and Science · doi:10.56726/irjmets424650
Design and Implementation of Multi-mode Sensor Signal Processor on FPGA Device
Kang, S., & Jung, Y. (2023) · Journal of Sensor Science and Technology · doi:10.46670/jsst.2023.32.4.246
“IoT 시스템과 스마트 가전제품 등에 적용되는 센서 기반 엣 지 디바이스들은 편리함을 위해 실시간 처리와 저전력 설계가 요구된다 [7].”
Design and Implementation of Tiny ML Model Using STM32F Platform
Namratha, S., Bhagya, R., & Bharthi, R. (2024) · Algorithms for intelligent systems · doi:10.1007/978-981-99-9436-6_12 · Cited by 1
Yoo, S., Jung, J. H., Lee, J.-K., et al. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4762162
Embedding Artificial Intelligence into Wearable IoMT Systems
Puckett, S. C., Menon, V., & Jovanov, E. (2024) · Lecture notes in networks and systems · doi:10.1007/978-3-031-60840-7_27
RedgeX: Meta-Learning based Optimal Analytical Model for Programmable Edge Intelligence
Alam, M., Ahmed, N., Matam, R., & Barbhuiya, F. A. (2024) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc57260.2024.10570981
“Having a resource-optimized analytical model in edge device intelligence is essential because edge devices typically have limited resources in terms of processing power, memory, and energy [7].”
BackgroundTiny-Ml Model for Pugilism Sport Gesture Classification and Its Potential over Computer Vision
Sahasra, K., Jain, A., Savaridassan, P., & Likhitha, P. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-68905-5_41
Enabling On-Device Inference of Large Language Models : Challenges, Techniques, and Applications
Ramkumar, A. (2024) · International Journal of Scientific Research in Computer Science Engineering and Information Technology · doi:10.32628/cseit241061100
Ultralow power communication: Challenges and machine learning solutions
Sharma, R., Singh, B., Prashar, D., & Jha, S. (2025) · Elsevier eBooks · doi:10.1016/b978-0-443-29927-8.00012-6
Energy Efficient Neural Architectures for TinyML Applications
Faheem, M. (2025) · International Journal of Scientific Research and Modern Technology. · doi:10.38124/ijsrmt.v4i5.531
TinyML Applications in Micronutrient Sensing: A Review of Microcontroller DeploymentsPreprint
Moeketsi, D., Mkhantshwa, A., Modise, C., & Mlangeni, N. (2025) · Research Square · doi:10.21203/rs.3.rs-6844555/v1 · Cited by 2
Detecting Energy Attacks in the Battery-Less Internet of Things
Mottola, L., & Voigt, T. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-87496-3_11 · Cited by 1
Alharthi, R. S., Alhussein, M. A., Fareed, K., Islam, M., & Aurangzeb, K. (2025) · International Journal of Imaging Systems and Technology · doi:10.1002/ima.70145 · Cited by 1
TinyML Empowering Intelligent Edge Devices
Joy, H. K., Jayarani, E. A., & Sridevi, R. (2025) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-0746-6.ch003
TinyML for Low-Power Embedded Intelligent Systems: A ReviewPreprint
Bourekouche, H. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5431335 · Cited by 11
Towards a Distributed Quantized Machine Learning Inference Using Commodity SoC-FPGAs Using FINN
Hannoun, M., Zuckerman, S., & Romain, O. (2025) · WiPiEC Journal - Works in Progress in Embedded Computing Journal · doi:10.64552/wipiec.v11i1.76
He, C., Wang, C., Jiang, W., Chen, T., Zhang, S., & Zhang, Y. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5492548
Time Series-Based Electrical Device Classification on Edge with TinyML
Reis, T., & Naskalı, A. T. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-05994-9_7 · Cited by 1
Velocity Estimation in IoT and Body Area Networks: A Review
Moussa, K. H., El-Sakka, A. H., & Soliman, M. A. (2025) · International Telecommunications Conference (ITC-Egypt) · doi:10.1109/itc-egypt66095.2025.11186669
TinyML Approach for Pre-fall Motion Pattern Detection in Older Adults
Sarmiento-Rojas, J., Torres-Lara, A. M., Ayala‐Parra, P., Jaramillo-Isaza, S., & Perdómo, O. (2025) · Communications in computer and information science · doi:10.1007/978-3-032-08203-9_13
AI-powered ambient assisted living
Bhattacharya, J., & Rakshit, P. (2025) · Elsevier eBooks · doi:10.1016/b978-0-443-14061-7.00009-9
Emerging Trends and Design Strategies in Low Power Electronics for Sustainable Systems
Shermy, R. P., Saranya, N., Akshaya, R., et al. (2025) · International Conference on Electronics and Sustainable Communication Systems (ICESC) · doi:10.1109/icesc65114.2025.11212432
Privacy-Preserving Electricity Load Forecasting in Smart Cities Using TinyML at the Edge
Almaini, A. E. A., Nataraj, S., Savaliya, G., et al. (2025) · IEEE International Conference on High Performance Computing and Communications (HPCC) · doi:10.1109/hpcc67675.2025.00152
“TinyML allows lightweight recurrent models to operate directly on edge devices such as smart meters [5], facilitating local inference without transmitting raw data.”
Enhancing Mobile Threat Detection through TinyML and Auditable ML
Majdoubi, C., Mendili, S. E., & Gahi, Y. (2025) · IEEE Congress on Information Science and Technology (CiSt) · doi:10.1109/cist65886.2025.11224069 · Cited by 1
Putra, A. N., Aradea, A., Rahmatulloh, A., Darmawan, I., Rizal, R., & Sutriani, L. (2025) · IEEE International Conference on Recent Advances and Innovations in Engineering (ICRAIE) · doi:10.1109/icraie65839.2025.11239099
“However, their computational complexity raises substantial environmental concerns [3], as training and inference require GPU/TPU resources with high energy consumption - particularly problematic for resource-constrained edge devices and IoT systems [4].”
An Unsupervised TinyML Approach with Efficient Edge AI for Effective Stress and Sleep Monitoring
Sudha, M., Neelaveni, P., Pooja, E., Mangaiyarkarasi, V., Begum, R., & Jayanthi, J. G. (2025) · International Conference on Recent Innovation in Science Engineering and Technology (ICRISET) · doi:10.1109/icriset64803.2025.11252416
Automated Access Control Mechanism for Industrial Floor Entry Using Face Recognition
Kumbhar, S., Jagtap, S., Bhamare, A., et al. (2025) · International Conference on Computing Communication Control and automation · doi:10.1109/iccubea65967.2025.11283944
“Tiny Machine Learning (TinyML) overcomes these issues by locally processing data and inferring models in microcontrollers with constrained resources[29].”
Maritime UAV Mission Classification and Pattern Analysis Based on Edge Intelligence Framework
Wu, J., Qiu, H., Qu, L., Liu, P., Li, N., & Wu, Y. (2025) · IEEE International Conference on Unmanned Systems (ICUS) · doi:10.1109/icus66297.2025.11294516
“…the evolution of edge intelligence frameworks has gone through three stages: (1) Terminal Intelligence Stage: represented by embedded AI technologies such as TinyML, edge devices are able to perform simple AI reasoning tasks, which promotes the intelligent application of low-power devices [6].”
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
Majdoubi, C., Mendili, S. E., Gahi, Y., & El‐Khatib, K. (2025) · Information · doi:10.3390/info17010021 · Cited by 1
Towards Efficient IoUT Vision: Simulated Compressed Image Transmission and Detection
Garikapati, J. S., Bhushan, C. M., Gazi, F., & Hussain, M. M. (2025) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf66845.2025.11317199
“The remainder of this paper is organized as follows: Section II surveys prior work in underwater vision [4] and communication.”
How to Train Your Metamorphic Deep Neural Network
Sommariva, T., Calderara, S., & Porrello, A. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-10192-1_25
WIP: tinyML Systems 101 New Course Curriculum in tiny Machine Learning
Aygün, S. (2025) · 2022 IEEE Frontiers in Education Conference (FIE) · doi:10.1109/fie63693.2025.11328713
“Thus, ML systems constitute an interdisciplinary field requiring knowledge from embedded systems, computer architecture, signal processing, and ML [8], [9].”
On-Device Artificial Intelligence for Image Analysis in Resource-Limited Settings: A Review
Xie, C., An, J. E., Bae, H.-J., Yun, E. S., Ko, K.-C., Jeong, D. S., & Kwon, O. S. (2025) · Applied Science and Convergence Technology · doi:10.5757/asct.2026.35.1.1 · Cited by 2
TinyML based monitoring and diagnosis system to enable the European battery digital passport
Ogliari, E., Mussetta, M., Carratù, M., Paciello, V., Pietrosanto, A., Pafumi, E., Ferrigno, L., & Milano, F. (2025) · Asia-Pacific Power and Energy Engineering Conference (APPEEC) · doi:10.1109/appeec66370.2025.11382158
“TinyML offers a promising approach by embedding lightweight neural networks into low-power microcontrollers, enabling real-time device monitoring and diagnostics [14].”
Integrating Machine Learning, Deep Learning, Iot, Ai, and Data Science for Future Innovations
Megha, & Srivastava, V. K. (2025) · International Conference on Innovations and Emerging Technologies in AI & Communication Systems (IETACS) · doi:10.1109/ietacs68750.2025.11385644
“In the article by Alajlan and Ibrahim (2022), the inference of running DL model on IoT edge gadgets that do not consume large amounts of energy is discussed [17].”
Subramanian., M., Subbulakshmi, A., S, S. B., Navin Athreya, R., Adithya, R., & Chezhiyan., M. (2025) · International Conference on Innovations and Emerging Technologies in AI & Communication Systems (IETACS) · doi:10.1109/ietacs68750.2025.11385778
“Such results motivate coupling structural design improvements with real time loading control to maximize both safety and efficiency[19].”
SteadyFlow-Edge: Adaptive Frame-Dropping for Stable Real-Time Detection on Edge
He, S.-R., Chen, C., Zheng, Q., et al. (2025) · International Conference on Advances in Computing and Artificial Intelligence · doi:10.1109/acai68217.2025.11406678
“With the continuous development of edge computing and embedded systems, the demand for edge inference systems in practical applications has been steadily increasing [1] [2].”
Comparative study of TinyML approaches for edge deployment in water quality
Hentati, M., & Hentati, R. (2026) · Journal of Environmental Engineering and Science · doi:10.1680/jenes.25.00112
Salem, M. A., Igathinathane, C., Pease, A. B., et al. (2026) · Journal of Agriculture and Food Research · doi:10.1016/j.jafr.2026.102899
Kumar, B. A., Hariharan, G., & Morin, L. S. (2026) · doi:10.1109/icvadv67766.2026.11469879
ML Guided DVFS with Cipher Selection for Secure, Low-Power VLSI IoT Devices
Arif, R., Mariammal, K., Durkadevi, C., Selvarani, N., Sakthivel, B., & Mahalakshmi, S. (2026) · International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES) · doi:10.1109/icses66558.2026.11479026
Low-Latency ML Offloading Across Edge and IoT Devices
Papazafeiropoulos, K., Mallikopoulou, A., Nanos, A., et al. (2026) · ACM/SPEC International Conference on Performance Engineering (ICPE) · doi:10.1145/3777884.3796989
“These applications rely on low-latency decision-making from complex sensor data, often on devices with severe resource constraints such as microcontrollers or edge nodes [1].”
Lightweight Deep Learning Models for Real-Time Edge Intrusion Detection
Islam, M. S., Muntaha, S., Salma, S., et al. (2025) · International Conference on Computer and Information Technology (ICCIT) · doi:10.1109/iccit68739.2025.11490486
“It enables low-latency decision-making, improved energy efficiency and enhanced privacy by keeping data local [11], [12].”
Bhushan, C. M., Koppuravuri, P., Prasanthi, N., Gazi, F., & Hussain, M. M. (2026) · International Conference on Computational Intelligence and Networks (CINE) · doi:10.1109/cine68769.2026.11503014
“It reduces latency, bandwidth consumption, and security issues when compared to sending data to the cloud for machine learning computations [14].”
Fernandez, S. M. B. (2026) · International Conference on Inventive Computation Technologies (ICICT) · doi:10.1109/icict68280.2026.11510813
Wearable systems for real-time disease diagnosis and predictive analytics
Huynh, P. K., Nguyen, D. H., Liu, H., Marahatta, S., Olaniran, O., Nguyen, D., Truong, T. T., Huynh, P., Keshwani, A., Prioli, J. P. J., et al. (2026) · Elsevier eBooks · doi:10.1016/b978-0-44-331426-1.00018-8
On-Device TinyML-Enabled Multimodal Threat Detection for Women and Child Safety Applications
Adiyattil, R., Jahnavi, K. S., Pooja.R, et al. (2026) · doi:10.1109/iccrtee68719.2026.11566199
“In recent times, AI has seen the development of TinyML and edge artificial intelligence (Edge AI) that is capable of running intelligent data processing directly on low-power microcontrollers [5].”
Adebayo, P. O., Ojo, O. S., Olatunji, M. A., et al. (2026) · Elsevier eBooks · doi:10.1016/b978-0-443-27692-7.00019-8
Gozuoglu, A., & Elhattab, M. (2026) · Bitlis Eren Üniversitesi Fen Bilimleri Dergisi · doi:10.17798/bitlisfen.1748636
Alanazi, A. G., Alanazi, H. A., & Albalawi, N. S. (2026) · Electronics · doi:10.3390/electronics15132918 · Cited by 1
Towards Human-Centric Network Operations: Leveraging Large Language Models as Unified Interfaces
Conde-Acereda, D., Fondo‐Ferreiro, P., & Gil‐Castiñeira, F. (2026) · International Conference on Communications, Signal Processing, and their Applications (ICCSPA) · doi:10.1109/iccspa69228.2026.11600868
“4, is considered the most efficient, as it involves fine-tuning lightweight LLMs, known as Tiny Language Models [15], to understand the specific configuration language of network devices without needing to handle large volumes of general knowledge.”
Hybrid Edge–Cloud Asymmetric Analytics for Portable Multimodal BCI Biosensors
Ghosh, S., Sindhujaa, P., Kumar, P. S., Mohan, A., Mahalakshmi, P., Gulyás, B. Z., Máthé, D., & Padmanabhan, P. (2026) · Biosensors · doi:10.3390/bios16070394 · Cited by 2
“The edge tier is reserved for time-bounded operations, namely windowed segmentation, schema-constrained feature generation, and lightweight fixed-parameter inference, all of which must execute within deterministic resource budgets and without dependence on continuous network availability [53].”
A Lightweight 1D CNN Framework for UAV Acoustic Detection in Complex Noise Environments Using TinyML
Wang, Z. (2026) · IEEE International Conference on Electronic Technology, Communication and Information (ICETCI) · doi:10.1109/icetci68937.2026.11610810
“However, these recurrent models often suffer from high computational latency and complex memory access patterns, making them sub-optimal for ultra-low-power microcontrollers [7, 8].”
Rokaya, M., AlZain, M. A., Hemdan, D. I. I., Atwa, M. A., & Atlam, E.-S. (2026) · Journal of Disability Research · doi:10.57197/jdr-2026-0888
Intelligent Distributed Communication Routing in IoT Networks Using Predictive Local AI Models
Alrashidi, M., & Abdallah, W. (2026) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs69761.2026.11613501
Attention in Motion: Secure Platooning via Transformer-Based Misbehavior Detection
Kalogiannis, K., Hussain, A., Li, H., & Papadimitratos, P. (2026) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2026.3711275 · Cited by 3
“Long et al. [66] applied transformers to cloud network intrusion detection, achieving 97.3% accuracy with significantly reduced false positive rates compared to traditional DNN approaches.”
InfluentialMadhalingam, H., Ramesh, N. M., Jaffarullah, A. A., Shanmugavel, B., Mathi, S., & Abdullahi, A. M. (2026) · IEEE Access · doi:10.1109/access.2026.3721366
A Comprehensive Survey on Embedded Deep Learning Algorithms for Smart IoT Monitoring Systems
Alalwany, W. S. H., & Fahim, A. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-25568-6_9
Guan, S., Liao, Z., Han, X., & Niu, S. (2026) · iScience · doi:10.1016/j.isci.2026.116422 · Cited by 1
Metadata
- Authors
- Alajlan, N. N., & Ibrahim, D. M.
- Affiliations
- Qassim University (SA)
- Corresponding author
- Ibrahim, Dina M.
- Venue
- Micromachines
- Publisher
- MDPI AG
- Published
- 2022-05-29
- Volume
- 13
- Issue
- 6
- Pages
- 851
- ISSN
- 2072-666X (ISSN-L)
- PMC
- PMC9227753
- OpenAlex
- W4282025104
- Cited by
- 224
- References
- 37
- Citations by year
- 2026: 67 · 2025: 71 · 2024: 56 · 2023: 26 · 2022: 4
- FWCI
- 17.41
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Micromachines (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Information Systems › Internet of Things and AI
- Keywords
- TinyML, edge computing, Internet of Things, ultra-low power electronics, microcontroller, deep learning inference, on-device learning, resource-constrained devices, IoT datasets
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
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.