Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Fine-tuned convolutional networks can classify lung X-rays with high accuracy, and the paper argues ResNet-50v2 is the best overall despite VGG16's highest raw score.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 12:55 UTC pith:HBE7HGRA

load-bearing objection Claims four-class lung-disease classification but only trains on three COVID-era classes; the stated contribution is untested and the 'ResNet-50 best' conclusion is contradicted by the paper's own Table 2. the 4 major comments →

arxiv 2607.26580 v1 pith:HBE7HGRA submitted 2026-07-29 cs.CV cs.AIcs.LG

Classification of Disease from Lungs X-ray Images using VGG16, VGG19 and ResNet50 Models

Nand Lal Yadav , Rajesh Kumar , Satyendra Singh , Sudhakar Singh This is my paper
classification cs.CV cs.AIcs.LG
keywords chest X-raytransfer learningVGG16VGG19ResNet-50lung disease classificationimage augmentationmedical image analysis
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that transfer learning with standard convolutional networks — VGG16, VGG19, and ResNet-50v2 — can classify chest X-rays into lung-disease categories, and that ResNet-50v2 is the strongest of the three. The experiments compare the models with and without image augmentation, reporting accuracies between 85% and 96%; the tables put VGG16 without augmentation at 96.07% as the single highest figure, with augmented ResNet-50v2 at 95% as the best of the augmented runs. A sympathetic reading of the central claim is that off-the-shelf pretrained CNNs, fine-tuned on a relatively small X-ray set, are accurate enough to support computer-aided pulmonary screening. The paper frames the task as pneumonia, tuberculosis, lung cancer, and normal lungs, but its dataset section describes only Covid, Normal, and Viral Pneumonia images. That gap between the stated disease scope and the actual three-class data is where the conclusion overreaches.

Core claim

On the paper's own terms, VGG16, VGG19, and ResNet-50v2 are each pretrained for general image recognition, fine-tuned on a public chest X-ray dataset, and tested on Covid, Normal, and Viral Pneumonia images. Without augmentation, VGG16 reaches 96.07%, VGG19 92.49%, and ResNet-50v2 93%; with augmentation, the results are 94%, 85%, and 95%, respectively. The paper interprets the ResNet-50v2 result as evidence that its residual architecture is best at capturing complex radiological features, and concludes that these models can be implemented in pulmonary disease diagnosis for early detection. The tables themselves show VGG16 without augmentation holding the highest raw accuracy, so the "ResNet-

What carries the argument

The machinery is transfer learning on three named CNN backbones: VGG16 (13 convolutional layers plus 3 fully connected layers), VGG19 (16 convolutional layers plus 3 fully connected layers), and ResNet-50v2 (50 layers with residual skip connections and global average pooling). The skip connections in ResNet-50v2 are the reason the paper gives for its high accuracy, since they let gradients reach deep layers during training. On the data side, a standard augmentation pipeline — rescaling, sample centering, rotation, width/height shifts, shear, zoom, and horizontal flip — expands the small X-ray set. The machinery's job is to convert general-purpose pretrained features into a three-way lung X-r

Load-bearing premise

The central claim rests on the assumption that the described three-class dataset (Covid, Normal, Viral Pneumonia) and the unreported training protocol justify a four-disease lung-screening conclusion; if the dataset contains no tuberculosis or lung cancer images, that scope is unsupported.

What would settle it

Download the cited chest X-ray dataset and list every class directory: it contains only Covid, Normal, and Viral Pneumonia, so no confusion matrix from any model trained as described can include tuberculosis or lung cancer; this directly tests the four-disease claim. A reproduction with a fixed seed, stated split, and reported epoch/optimizer settings would also show whether ResNet-50v2's 95% accuracy is stable or depends on the unreported protocol.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the reported accuracies hold, a three-way Covid/Normal/Viral-Pneumonia screening tool on this X-ray dataset can exceed 95% accuracy with an augmented ResNet-50v2.
  • Fine-tuning pretrained CNNs could let clinics with modest imaging datasets deploy automated X-ray triage without training models from scratch.
  • The comparison yields a concrete architecture choice: use ResNet-50v2 with augmentation when working with chest X-rays among these three models.
  • Augmentation does not help uniformly — VGG16 and VGG19 accuracy drops under augmentation while ResNet-50v2 improves — so preprocessing choice matters as much as model choice.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The abstract and conclusion speak of pneumonia, tuberculosis, and lung cancer, but the experiments described cover only Covid, Normal, and Viral Pneumonia; I read that as meaning the paper's evidence does not support a TB or lung-cancer accuracy claim.
  • The training protocol — epochs, optimizer, learning rate, batch size, and exact train/test split — is never stated, so the headline accuracies are tied to unreported settings; a fixed protocol could change the model ranking.
  • A natural extension would be to run the same three-architecture comparison on a larger chest X-ray dataset that actually contains tuberculosis, lung cancer, and other pathologies, reporting per-class sensitivity rather than a single accuracy.
  • Because augmentation helped only ResNet-50v2, the augmentation parameters may need to be tuned per architecture, or the three models may be at different points along the training/validation trade-off.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper compares three transfer-learning CNN architectures — VGG16, VGG19, and ResNet-50v2 — for chest X-ray classification on the Kaggle 'covid19-image-dataset', with and without data augmentation. It reports per-class precision, recall, F1, accuracy, loss, and confusion matrices. The abstract and introduction, however, state that the models classify 'pneumonia, tuberculosis, lung cancer, and normal lungs' and conclude that ResNet-50 performs best. The body of the paper reports only three classes (Covid, Normal, Viral Pneumonia), and the dataset description in Section 3 mentions no tuberculosis or lung cancer images. The experimental protocol (class counts, split sizes, optimizer, learning rate, batch size, epoch count, seed, and test-set usage) is not reported.

Significance. If the stated four-class claim were supported, this would be a modest but potentially useful empirical comparison of off-the-shelf CNNs for multiclass lung-disease screening. The paper has some strengths: it uses an external public dataset rather than a self-collected one, reports per-class metrics beyond overall accuracy, and provides confusion matrices. However, the central claim as stated is not tested: no experiment involves tuberculosis or lung cancer. The ResNet-50 superiority claim is also internally inconsistent across the two tables. Consequently, the significance of the actual experimental contribution is limited and does not support the abstract's conclusions.

major comments (4)
  1. [Abstract; §1; §3; Tables 2–3] The abstract and introduction claim classification of pneumonia, tuberculosis, lung cancer, and normal lungs. However, §3 states that the dataset contains only 'Covid,' 'Normal,' and 'Viral Pneumonia,' and Tables 2–3 report metrics for exactly these three classes. No tuberculosis or lung cancer images appear anywhere in the paper. The central claim of the paper is therefore untested by the reported experiments.
  2. [§3.1; Tables 2–3] The experimental protocol is critically underreported. The paper gives augmentation parameters and validation_split=0.2, but no class counts, train/test split sizes, diagnosis of whether the original Kaggle test split was used, optimizer, learning rate, batch size, number of epochs, loss function, or random seeds. Combined with the small accuracy differences between models (e.g., 0.95 vs. 0.94 in Table 3), the reported results cannot be reproduced or assessed for statistical significance.
  3. [§5; Tables 2–3] The conclusion that 'ResNet-50 performs best' is not supported by the reported numbers. Without augmentation, VGG16 has the highest accuracy (0.9607 in Table 2) compared to ResNet-50v2 (0.93), and ResNet-50v2 only leads by 0.01 (0.95 vs. 0.94) with augmentation in Table 3, with no variance, confidence intervals, or multiple-seed results. The claim of ResNet-50 superiority is therefore internally inconsistent with the authors' own tables.
  4. [§1; §3] The paper's framing implies that a 'vast amount of X-ray images' covering multiple pathologies was used, and the literature review refers to datasets with 'pneumonia, tuberculosis, lung cancer, and normal lung conditions.' The actual dataset described in §3 is the small Kaggle COVID-19 chest X-ray dataset. The mismatch between the claimed scope and the actual data should be corrected or the claims substantially narrowed.
minor comments (4)
  1. [Throughout] Model naming is inconsistent: the abstract and title say 'ResNet50,' while the body uses 'ResNet-50v2.' Clarify which architecture was actually used.
  2. [References] References [13] and [17] are the same paper (He et al., deep residual learning) and should be merged. Also, reference [14] CheXpert is cited as a dataset containing tuberculosis and lung cancer, but CheXpert does not provide those labels; check the citation.
  3. [§3.3] The definitions of TP/TN/FP/FN are imprecise ('predicted positively' is not enough; specify the actual class). Sensitivity/specificity are defined in §3.3.3 but never reported in any table.
  4. [§4] Figures 2–7 plot training/validation accuracy and loss, but the curves are too small to read in the provided text; this limits their usefulness. Higher-resolution figures or numerical summaries would help.

Circularity Check

0 steps flagged

No circularity: reported accuracies are direct empirical measurements, not derivations from their own inputs.

full rationale

The paper's results are obtained by training standard, externally pretrained CNN architectures (VGG16, VGG19, ResNet-50v2) on a public Kaggle dataset and measuring accuracy, precision, recall, and F1 on held-out test images. There is no derivation chain in which a predicted quantity is defined in terms of the fitted quantity, no fitted parameter is renamed as a prediction, and no load-bearing self-citation is used to force the conclusion. The main problems with the paper are substantive correctness concerns rather than circularity: the dataset in Section 3 contains only Covid, Normal, and Viral Pneumonia images, so the abstract's claim about classifying tuberculosis and lung cancer is untested; and the conclusion that ResNet-50 performs best conflicts with Table 2, where VGG16 has the highest accuracy. These are evidence/validity issues, not cases where the result reduces to its own inputs by construction. Accordingly, the circularity score is 0.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

This is an empirical benchmark rather than a derivation, so the ledger captures the experiment's implicit purchases: public labels taken as truth, ImageNet transfer assumed to be appropriate, unreported training hyperparameters, and hand-set augmentation settings that change the headline numbers by up to 8 points (VGG19: 92.49→85%). The gap between the abstract's stated disease scope (TB, lung cancer) and the dataset's actual classes is a scope mismatch, not an axiom.

free parameters (3)
  • augmentation hyperparameters (rotation_range=40, width/height_shift_range=0.2, shear_range=0.2, zoom_range=0.2, horizont = stated in §3.1
    Hand-chosen standard augmentation settings; no tuning or sensitivity analysis is reported, and they materially change the reported accuracies (VGG16 96.07→94%, VGG19 92.49→85%, ResNet-50v2 93→95%).
  • validation_split = 0.2
    Hand-chosen fraction held out for validation; the paper does not state how this interacts with the original Kaggle train/test split.
  • training hyperparameters (optimizer, learning rate, batch size, epochs, loss) = not reported
    None are stated anywhere in §3 or §4, yet the reported accuracies, loss curves, and confusion matrices depend on them; any replication must guess them.
axioms (4)
  • domain assumption ImageNet-pretrained VGG16/VGG19/ResNet-50v2 weights transfer to chest X-ray classification (§3.2, implicit in the whole method)
    The paper fine-tunes models initialized from general image classification without testing whether this prior is appropriate for X-ray data; the claim depends on this transferability.
  • domain assumption The Kaggle dataset's labels (Covid / Normal / Viral Pneumonia) are clinically correct ground truth (§3)
    No radiologist verification or label audit is reported; the central accuracy numbers inherit whatever noise is in the public labels.
  • domain assumption The augmentation/validation protocol does not leak training data into validation or test (§3.1)
    validation_split=0.2 is applied inside the Keras ImageDataGenerator flow, but no code is shown; the paper never states which generator feeds validation, so leakage from augmented training samples into validation cannot be ruled out.
  • standard math VGG16/VGG19/ResNet-50v2 are implemented as described in [12], [13], [22]
    The architectures are taken as given from prior literature; this is unproblematic background but is still a reliance on external implementations.

pith-pipeline@v1.3.0-daily-deepseek · 8972 in / 16698 out tokens · 149316 ms · 2026-08-01T12:55:52.431408+00:00 · methodology

0 comments
read the original abstract

With the increase in the number of cases related to respiratory diseases, there is an urgent need to detect them early and diagnose them accurately. Convolutional neural networks have given promising results when used for diagnosing diseases using imaging tests. In this study, we investigate the potential of applying deep learning algorithms such as VGG16, VGG19, and ResNet50 for classification of lung ailments based on X-ray images. A detailed analysis of the aforementioned models' performances was conducted to assess how well they can classify various types of lung ailments, including pneumonia, tuberculosis, lung cancer, and normal lungs. In order to do that, these deep learning models were trained on a vast amount of X-ray images. The results of our study show that while all three models provide good results, ResNet-50 performs best in comparison with other models due to its efficiency and high level of accuracy. We believe that these deep learning models can be successfully implemented in the practice of diagnosing pulmonary diseases in the future. It helps with early disease detection and improves patient outcomes.

Figures

Figures reproduced from arXiv: 2607.26580 by Nand Lal Yadav, Rajesh Kumar, Satyendra Singh, Sudhakar Singh.

Figure 5
Figure 5. Figure 5: Performance variation of the VGG16 model with augmentation [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Performance variation of the VGG19 model with augmentation [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Performance variation of the ResNet50v2model with augmentation The results of the multiclass classifier models have been observed from the experimental analysis, and the multiclass classifier’s confusion matrix is shown in figures [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

41 extracted references · 2 linked inside Pith

  1. [1]

    Lung Disease Disorder refers to the wide range of disorders related to but not limited to pneumonia, tuberculosis (TB), chronic obstructive lung disease (COPD), and lung cancer

    INTRODUCTION There is an enormous number of deaths and sickness related to the lung disease, and it is a global health problem. Lung Disease Disorder refers to the wide range of disorders related to but not limited to pneumonia, tuberculosis (TB), chronic obstructive lung disease (COPD), and lung cancer. Lung diseases include different kinds of lung disea...

  2. [2]

    Pneumonia Infecting this disease leads to inflammation in the air sacs of the lungs that have the symptoms of fever, cough, and difficulty in breathing

  3. [3]

    Its symptoms include chronic coughing, chest pains, and weight loss

    Tuberculosis (TB) The TB disease is a bacterial infection which involves the lungs and is caused by Mycobacterium tuberculosis. Its symptoms include chronic coughing, chest pains, and weight loss

  4. [4]

    Its symptoms include coughing, blood-streaked phlegm, and chest pain

    Lung Cancer This is a type of cancer where abnormal cells form in the lung and its most common cause is smoking. Its symptoms include coughing, blood-streaked phlegm, and chest pain

  5. [5]

    This causes airway irritation that causes reduced breath and chronic coughing

    Chronic Obstructive Pulmonary Disease (COPD) Emphysema and chronic bronchitis are some examples of conditions that are included in the Chronic Obstructive Pulmonary Disease. This causes airway irritation that causes reduced breath and chronic coughing

  6. [6]

    Asthma Asthma is a chronic respiratory illness characterized by airway irritation and increased reactivity, resulting in recurrent episodes of wheezing, coughing, and breathlessness

  7. [7]

    Interstitial Lung Disease (ILD) ILD encompasses a group of disorders which create scarring of the lung tissue, affecting its ability to expand and contract properly

  8. [8]

    Pulmonary Hypertension This condition involves increased blood pressure in the pulmonary arteries, showing symptoms like shortness of breath, tiredness, and chest pain

  9. [9]

    Cystic Fibrosis Cystic fibrosis is a genetic disorder harming the lungs and other organs, leading to the growth of thick and sticky mucus, causing respiratory issues

  10. [10]

    Lung Fibrosis Lung fibrosis refers to the scarring of lung tissue due to various causes, resulting in reduced lung function and breathlessness

  11. [11]

    It can be caused by various factors, such as a bacterial infection

    Lung Abscess Lung abscess is defined as a localized collection of pus that develops within the tissue of the lungs. It can be caused by various factors, such as a bacterial infection. Early diagnosis and effective treatment of respiratory diseases can be achieved through screening. It has been realized that the suitability of certain screening practices h...

  12. [12]

    Interpretation of X -ray images manually is one of the conventional methods of diagnosis, which is time consuming and prone to human error

    LITERATURE REVIEW The lung diseases that are a massive health burden to the world are pneumonia, TB and lung cancer. Interpretation of X -ray images manually is one of the conventional methods of diagnosis, which is time consuming and prone to human error. To counter this, the researchers have resorted to deep learning in ensuring that they create automat...

    2016

  13. [13]

    train" and

    MATERIALS AND METHODS The covid19-image-dataset1 is a collection of medical images related to respiratory conditions, particularly COVID-19. The dataset is divided into two subsets, i.e. "train" and "test". It 1https://www.kaggle.com/datasets/prashant268/chest-xray-covid19- pneumonia/download?datasetVersionNumber=2 includes images cate gorized into "Covid...

  14. [14]

    M., Adams, L

    Niehues, S. M., Adams, L. C., Gaudin, R. A., Erxleben, C., Keller, S., Makowski, M. R., ... &Bressem, K. K. (2021). Deep -learning-based diagnosis of bedside chest X -ray in intensive care and emergency medicine. Investigative Radiology, 56(8), 525-534

    2021

  15. [15]

    Gonem, S., Janssens, W., Das, N., & Topalovic, M. (2020). Applications of artificial intelligence and machine learning in respiratory medicine. Thorax, 75(8), 695-701

    2020

  16. [16]

    S., Divya, G

    Jogin, M., Madhulika, M. S., Divya, G. D., Meghana, R. K., & Apoorva, S. (2018, May). Feature extraction using convolution neural networks (CNN) and deep learning. In 2018 3rd IEEE international conference on recent trends in electronics, information & communic ation technology (RTEICT) (pp. 2319-2323). IEEE

    2018

  17. [17]

    (2023, January)

    Suryawanshi, V., Adivarekar, S., Bajaj, K., & Badami, R. (2023, January). Comparative Study of Regularization Techniques for VGG16, VGG19 and ResNet-50 for Plant Disease Detection. In International Conference on Communication and Computational Technologies (pp. 771 - 781). Singapore: Springer Nature Singapore

    2023

  18. [18]

    Sekine, Y., Katsura, H., Koh, E., Hiroshima, K., & Fujisawa, T. (2012). Early detection of COPD is important for lung cancer surveillance. European Respiratory Journal, 39(5), 1230 - 1240

    2012

  19. [19]

    F., Hurd, S., Anzueto, A., Barnes, P

    Rabe, K. F., Hurd, S., Anzueto, A., Barnes, P. J., Buist, S. A., Calverley, P., ... & Zielinski, J. (2007). Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: GOLD executive summary. American journal of respiratory and critical care medicine, 176(6), 532-555

    2007

  20. [20]

    E., Hemdan, E

    Karar, M. E., Hemdan, E. E. D., & Shouman, M. A. (2021). Cascaded deep learning classifiers for computer-aided diagnosis of COVID-19 and pneumonia diseases in X-ray scans. Complex & Intelligent Systems, 7, 235-247

    2021

  21. [21]

    Halalli, B., &Makandar, A. (2018). Computer aided diagnosis -medical image analysis techniques. Breast imaging, 85, 85-109

    2018

  22. [22]

    E., Setio, A

    Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., ... & Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical image analysis, 42, 60-88

    2017

  23. [23]

    LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. nature, 521(7553), 436-444

    2015

  24. [24]

    L., Zhu, K., Yang, B., Mehta, H.,

    Rajpurkar, P., Irvin, J., Ball, R. L., Zhu, K., Yang, B., Mehta, H., ... & Lungren, M. P. (2018). Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists. PLoS medicine, 15(11), e1002686

    2018

  25. [25]

    Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large -scale image recognition. arXiv preprint arXiv:1409.1556

    Pith/arXiv arXiv 2014

  26. [27]

    Irvin, J., Rajpurkar, P., Ko, M., Yu, Y., Ciurea-Ilcus, S., Chute, C., ... & Ng, A. Y. (2019, July). Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. In Proceedings of the AAAI conference on artificial intelligence (Vol. 33, No. 01, pp. 590-597)

    2019

  27. [28]

    A Deep -Learning-Based Framework for Automated Diagnosis of COVID-19 Using X-ray Images

    Khan, I.U.; Aslam, N. A Deep -Learning-Based Framework for Automated Diagnosis of COVID-19 Using X-ray Images. Information 2020, 11, 419

    2020

  28. [29]

    A., Ko, J., Swetter, S

    Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. nature, 542(7639), 115-118

    2017

  29. [30]

    He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778)

    2016

  30. [31]

    Alshmrani, G. M. M., Ni, Q., Jiang, R., Pervaiz, H., &Elshennawy , N. M. (2023). A deep learning architecture for multi -class lung diseases classification using chest X -ray (CXR) images. Alexandria Engineering Journal, 64, 923-935

    2023

  31. [32]

    (2016, September)

    Fawzi, A., Samulowitz, H., Turaga, D., & Frossard, P. (2016, September). Adaptive data augmentation for image classification. In 2016 IEEE international conference on image processing (ICIP) (pp. 3688-3692). Ieee

    2016

  32. [33]

    Yang, H., Ni, J., Gao, J., Han, Z., & Luan, T. (2021). A novel method for peanut variety identification and classification by Improved VGG16. Scientific Reports, 11(1), 15756

    2021

  33. [34]

    D., Rajinikanth, V., Pugalenthi, R., & Raja, N

    Dey, N., Zhang, Y. D., Rajinikanth, V., Pugalenthi, R., & Raja, N. S. M. (2021). Customized VGG19 architecture for pneumonia detection in chest X-rays. Pattern Recognition Letters, 143, 67-74

    2021

  34. [35]

    Prusty, S., Patnaik, S., & Dash, S. K. (2022, August). ResNet50V2: A Transfer Learning Model to Predict Pneumonia with chest X-ray images. In 2022 International Conference on Machine Learning, Computer Systems and Security (MLCSS) (pp. 208-213). IEEE

    2022

  35. [36]

    Halder, A., & Datta, B. (2021). COVID-19 detection from lung CT-scan images using transfer learning approach. Machine Learning: Science and Technology, 2(4), 045013

    2021

  36. [37]

    H., Khan, T., & Khan, S

    Syed, A. H., Khan, T., & Khan, S. A. (2023). Deep Transfer Learning Techniques -Based Automated Classification and Detection of Pulmonary Fibrosis from Chest CT Images. Processes, 11(2), 443

    2023

  37. [38]

    (2013, October)

    Koço, S., & Capponi, C. (2013, October). On multi -class classification through the minimization of the confusion matrix norm. In Asian Conference on Machine Learning (pp. 277-292). PMLR

    2013

  38. [39]

    Susmaga, R. (2004). Confusion matrix visualization. In Intelligent Information Processing and Web Mining: Proceedings of the International IIS: IIPWM ‘04 Conference held in Zakopane, Poland, May 17–20, 2004 (pp. 107-116). Berlin, Heidelberg: Springer Berlin Heidelberg

    2004

  39. [40]

    Hossin, M., & Sulaiman, M. N. (2015). A review on evaluation metrics for data classification evaluations. International journal of data mining & knowledge management process, 5(2), 1

    2015

  40. [41]

    A., Baloglu , U

    Ozturk, T., Talo, M., Yildirim, E. A., Baloglu , U. B., Yildirim, O., & Acharya, U. R. (2020). Automated detection of COVID -19 cases using deep neural networks with X -ray images. Computers in biology and medicine, 121, 103792

    2020

  41. [42]

    Metrics for multi -class classification: an overview

    Grandini M, Bagli E, Visani G. Metrics for multi -class classification: an overview. arXiv preprint arXiv:2008.05756. 2020 Aug 13

    Pith/arXiv arXiv 2008