CoChat
Sign Up
Cited by 70Advanced Neural Network ApplicationsFace recognition and analysisIoT and Edge/Fog Computing

A Tiny CNN Architecture for Medical Face Mask Detection for Resource-Constrained Endpoints

Open access
Mohan, Puranjay, Paul, Aditya Jyoti, Chirania, Abhay·2021·Lecture Notes in Electrical Engineering
DOI 10.1007/978-981-16-0749-3_52
Open this paper in CoChat — free accountOpen PDF

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1007/978-981-16-0749-3_52

Cites

17 of 17 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The Mohan et al. (2021) model is trained using TensorFlow and incurs 138 KB memory during post-quantization at 30 frames/s.

    “The model is trained by using the TensorFlow framework that incur 138 KB memory during post-quantization at 30 frames/s.”

    From A review on TinyML: State-of-the-art and prospects (2022) · cites this paper as Mohan et al., 2021

    Training Specifications (p. 5) says the models were designed in TensorFlow 2.3, and Our Proposed Model (p. 6) reports a reduction to 138 KB after quantization. The 138 KB is model size, not measured memory usage. The body passages found do not establish 30 frames/s; that figure appears only in the abstract, which is not evidence under the stated rubric. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • Mohan et al. (2021) deploy a convolutional neural network on an ARM Cortex-M7 MCU clocked at 480 MHz to perform face detection using 496 KB buffer memory.

    “For example, (Mohan et al., 2021) deploys a convolution neural network on ARM Cortex-M7 MCU clocked at 480 MHz to perform face detection by using 496 KB buffer memory.”

    From A review on TinyML: State-of-the-art and prospects (2022) · cites this paper as Mohan et al., 2021

    The paper’s opening passage (p. 1) reports deploying a model on an ARM Cortex-M7 board clocked at 480 MHz with 496 KB of framebuffer RAM. But the Introduction (p. 2) and Conclusion (p. 9) describe the authors’ CNN as detecting or recognizing face masks, not performing face detection; the 496 KB is a board memory specification, not a reported amount of memory used by the model. (LLM, conf=0.90)

    Experimental — accuracy not yet measured

  • In [29], after quantization, only the proposed model could deploy on the device, with 138 KB model size, 30 FPS inference speed, and 99.83% accuracy.

    “Overall, after performing the quantization technique on all three models to reduce the size, the results demonstrated that only the proposed model can deploy on the device, with 138 KB model size; the inference speed was 30 FPS and the accuracy was 99.83%.”

    From TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications (2022) · cites this paper as [29]

    Section 4.5 (p. 6) says only the proposed model could be evaluated on the OpenMV Cam; the two SqueezeNet models exceeded the device's 230 KB limit. Section 5.3 (p. 6) reports a 138 KB quantized model and 99.83% test accuracy, but the passages found do not report a 30 FPS inference speed. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Puranjay Mohan et al. [29] presented a tiny CNN model to detect Medical Face Masks in real-time on devices.

    “Puranjay Mohan et al. [29] presented a tiny CNN model to detect Medical Face Masks using real-time on devices.”

    From TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications (2022) · cites this paper as [29]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 5

  1. [3]3. Banner, R., Nahshan, Y., Soudry, D.: Post training 4-bit quantization of convolutional networks for rapid-deployment. In: H. Wallach, H. Larochelle, A. Beygelzimer, F. d ' Alch´ e-Buc, E. Fox, R. Garnett (eds.) Advances in Neural Information Processing Systems, vol. 32, pp. 7950-7958. Curran Associates, Inc. (2019). URL https://proceedings.neurips.cc/paper/2019/file/ c0a62e133894cdce435bcb4a5df1db2d-Paper.pdf
  2. [6]6. Dong, Z., Yao, Z., Cai, Y., Arfeen, D., Gholami, A., Mahoney, M.W., Keutzer, K.: Hawq-v2: Hessian aware trace-weighted quantization of neural networks. In: Advances in Neural Information Processing Systems 33 pre-proceedings (NeurIPS 2020) (2019)
  3. [15]15. LeCun, Y., Bottou, L., Bengio, Y., Haffner, P.: Gradient-based learning applied to document recognition. In: Intelligent Signal Processing, pp. 306-351. IEEE Press (2001)
  4. [20]20. Nahshan, Y., Chmiel, B., Baskin, C., Zheltonozhskii, E., Banner, R., Bronstein, A.M., Mendelson, A.: Loss Aware Post-training Quantization. arXiv e-prints arXiv:1911.07190 (2019) · arXiv
  5. [25]25. Zhao, R., Hu, Y., Dotzel, J., De Sa, C., Zhang, Z.: Improving neural network quantization without retraining using outlier channel splitting. In: K. Chaudhuri, R. Salakhutdinov (eds.) Proceedings of the 36th International Conference on Machine Learning, Proceedings of Machine Learning Research , vol. 97, pp. 7543-7552. PMLR, Long Beach, California, USA (2019). URL http://proceedings.mlr.press/ v97/zhao19c.html

Cited by

Sorted by citations

Related

Authors
Mohan, P., Paul, A. J., & Chirania, A.
Affiliations
SRM Institute of Science and Technology (IN)
Corresponding author
Mohan, Puranjay
ORCID
Venue
Lecture Notes in Electrical Engineering
Publisher
Springer Singapore
Published
2021
Pages
657-670
ISSN
  • 1876-1100 (ISSN-L)
  • 1876-1119
DOI
10.1007/978-981-16-0749-3_52
OpenAlex
W3110529155
Cited by
70
References
16
Citations by year
2025: 4 · 2024: 7 · 2023: 15 · 2022: 23 · 2021: 21
FWCI
24.36
Type
book chapter
Language
en
Open access
Licence
Locations
  • Lecture notes in electrical engineering (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
  • Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
Keywords
face mask detection, surgical masks, COVID-19 pandemic, TensorFlow Lite, model quantization, lightweight convolutional neural network, edge deployment, low memory footprint

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