Pith. sign in

REVIEW 4 major objections 6 minor 32 references

An Explorative Analysis of SVM Classifier and ResNet50 Architecture on African Food Classification

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read On six African dishes, a fine-tuned ResNet50 and a classical SVM both reach about 81% accuracy, with the SVM slightly ahead on macro F1.

desk verdict A modest, honest empirical comparison of SVM and ResNet50 on a small African foods dataset; the result is plausible, but the evaluation is thin on variance and the test set is unvalidated. read the letter →

arxiv 2505.13923 v1 pith:MND5FKMF submitted 2025-05-20 cs.CV

classification cs.CV
keywords AfricanfoodclassificationResNet50SupportVectorMachineTransferlearningimagerecognitionComparativeanalysisRawpixelfeaturesImbalanceddataset
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks a practical question: on a small, imbalanced dataset of six popular African dishes, does a deep convolutional network fine-tuned by transfer learning actually beat a classical SVM trained on raw pixel values? The authors report that both approaches reach about 81% overall accuracy on the same 251-image test set, with the SVM achieving a slightly higher macro-averaged F1 score (0.82 versus 0.81). The point is that on this dataset, a simple classical pipeline holds its ground against a modern deep model, while the two models differ in which classes they confuse. If the finding holds, it matters for food recognition in low-resource settings, where lightweight classical models are easier to deploy.

What carries the argument

The comparison rests on two pipelines. The SVM pipeline uses raw RGB pixel intensities resized to 100×100, flattened into 30,000-element vectors, and classified by an RBF-kernel SVM with default settings. The ResNet50 pipeline fine-tunes the last four layers of an ImageNet-pretrained ResNet50 with a global average pooling layer, a 1,024-neuron dense layer with dropout 0.5 and L2 regularization, and a 6-way softmax head, trained with 5-fold cross-validation and data augmentation. The 5-fold validation-based model selection and the fixed public test split are what make the two pipelines comparable.

What would settle it

Re-running the two pipelines on the same train/test split while excluding any test images that are near-duplicates of training images, or re-testing on an independently collected set of the same six dishes, would settle the claim: if accuracy of either model falls well below 81% under that check, the reported comparison is an artifact of the dataset split.

Watch

Extended reading notes

Core claim

The central claim is empirical and comparative. On the African Foods dataset of 1,658 images across six classes, a fine-tuned ResNet50 and a radial-basis-function SVM fed with flattened 100×100 RGB pixels both achieve an overall test accuracy of about 81%. Per-class results diverge: ResNet50 is strongest on Ndole (precision 1.00, F1 0.91) and Palm-nut soup (recall 0.90), while the SVM is more consistent, leading in Ekwang and Eru and posting the better macro F1 (0.82 vs 0.81). The paper reads these results as evidence that classical methods remain competitive on small, imbalanced food datasets and that per-class behaviour, not just overall accuracy, should drive model choice.

Load-bearing premise

The entire comparison is read from a single 251-image test split of the African Foods dataset; if those images are mislabeled or contain near-duplicates of training images, both 81% figures would be inflated and the two models could no longer be compared as reported.

Editorial extensions

If this is right

  • If the 81% figures are accurate, a lightweight SVM with raw pixels is a viable baseline for African food recognition on small datasets.
  • The macro F1 advantage of the SVM suggests that classical models can generalize more evenly across imbalanced classes in this setting.
  • ResNet50's strengths on particular classes imply that ensembles or hybrid pipelines could exploit the complementary errors of the two models.
  • The dataset-specific comparison indicates that without large, balanced data, deep transfer learning may not deliver a decisive advantage over a well-tuned classical method.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The result is specific to a 251-image test set; a natural next test is whether the SVM's edge survives on a larger, more balanced African food dataset, which the paper itself lists as future work.
  • The large confusion between palm-nut soup and ekwang suggests a visually similar pair that could be targeted with class-aware losses or hierarchical classification.
  • Because the SVM uses only raw pixels, adding handcrafted features or color histograms might improve it further, a direction the paper does not test.
  • The paper's reliance on a single public dataset means the 81% figure should be read as a benchmark for this dataset, not a claim about African food recognition in general.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. This paper presents an empirical comparison between a fine-tuned ResNet50 and a default RBF-kernel SVM using raw RGB pixel features for classifying six African food classes from a public Mendeley dataset of 1,658 images. The authors combine the original training and validation splits, use five-fold cross-validation for the ResNet50 and select the fold with the highest validation accuracy, train the SVM on the same combined training set, and evaluate both models on the original 251-image test split. They report overall accuracy of about 81% for both models, with the SVM obtaining a slightly higher macro-averaged F1 score (0.82 versus 0.81), and they conclude that a classical model can remain competitive with a transfer-learned deep network on this small dataset. The paper also provides per-class precision, recall, F1, and confusion matrices, and it releases code for the ResNet50 pipeline.

Significance. If the empirical claims hold, this is a modest but useful benchmark for African food recognition. The explicit comparison of a raw-pixel SVM with a fine-tuned CNN on a public African foods dataset is a data point that the community can build on, and the paper has the concrete strengths of using a publicly available dataset, reporting per-class metrics rather than only overall accuracy, and releasing code for the deep learning pipeline. However, the central comparative claim currently rests on a single small test split, a single selected validation fold, and an untuned SVM baseline, so the quantitative conclusion that SVM is more consistent than ResNet50 is not yet firmly established.

major comments (4)
  1. [Section II-A and Table III] The ResNet50 result is obtained from the single fold with the highest validation accuracy among five folds, rather than from repeated runs, averaged folds, or an ensemble. Selecting the best validation fold is an additional model-selection step on validation noise, so the reported 81% test accuracy does not represent the expected performance of the fine-tuning procedure. Please report per-fold test results and the mean plus/minus standard deviation, or use an ensemble over the folds, and also run multiple random seeds.
  2. [Section III-C and Table III] The headline comparative claim that the SVM is more consistent than ResNet50 rests on a macro-F1 difference of 0.01 (0.82 versus 0.81) and equal overall accuracy on a single 251-image test set. With per-class test sizes of 31 to 59 images, one or two misclassifications change a class-level F1 score by roughly 0.02 to 0.03, so the observed gap is within the magnitude of sampling noise. Bootstrap confidence intervals or a significance test are needed before concluding that one model is more consistent.
  3. [Section II-B and Section III-C] The SVM is trained with default scikit-learn hyperparameters (C=1.0, gamma='scale'), while the ResNet50 pipeline was manually tuned through a hyperparameter-selection process. This comparison confounds the model family with the amount of tuning effort. A simple grid search over C and gamma, or an explicit framing of the SVM as an untuned baseline, is needed before the paper can conclude that classical methods 'can still hold their ground' as a general alternative to deep learning.
  4. [Sections I-A, II-A, and II-B] All conclusions are read from the original 251-image test split, which the authors state they kept unchanged but did not audit. The paper reports no check for near-duplicate images across the training/test boundary, no label-noise analysis, and no representativeness analysis of the test images. Accidental leakage or systematic label errors would inflate both reported 81% accuracy figures, so a duplicate and near-duplicate check and, if feasible, a label audit should be added.
minor comments (6)
  1. [Abstract and text] There are several language and typographical issues: 'Food recognition systems has advanced' should be 'have advanced', 'african' should be capitalized, and 'Ghanian' should be 'Ghanaian'.
  2. [Section II-A] The phrase 'Randomly rotate images vertically (0.2)' is ambiguous; clarify whether this is a random rotation range or a different transformation, and distinguish it from the listed horizontal flip.
  3. [Section II-A] The hyperparameter selection is described only as 'manually selected' with no search ranges or number of trials, which limits reproducibility of the tuning process.
  4. [Contributions and Section II-A] The paper states that the full codebase is available, but only a GitHub link for the ResNet50 fine-tuning code is provided; no SVM or evaluation scripts are linked. Please add the missing code or adjust the claim.
  5. [References] Reference [18] is cited for the use of HOG features with SVMs, but the cited paper is titled 'Image classification using random forests and ferns'; please verify and correct this citation.
  6. [Figure 6] The plot shows average training and validation loss across five folds but does not show fold-to-fold variance; adding a shaded standard-deviation band would make the claim of stable convergence more informative.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's reported accuracies and F1-scores are empirical measurements on a held-out test split, with no fitted quantity renamed as a prediction.

full rationale

The paper contains no derivation chain whose conclusion is equivalent to its inputs. Both models are evaluated on a test set that was kept unchanged from the source dataset, while training used a combination of the original training and validation splits. The ResNet50 hyperparameters were selected manually based on validation accuracy during 5-fold cross-validation, and the SVM used default scikit-learn settings; neither selection feeds back into the definition of the reported metrics. The evaluation metrics (accuracy, precision, recall, F1, confusion matrix) are standard and computed directly from predictions on the 251-image test set. There are no fitted parameters that are then reinterpreted as predictions, no uniqueness theorem imported from prior work, and no ansatz smuggled in via citation. The two self-references involving the authors are incidental background citations for CNN concepts and residual learning, and they do not support the empirical comparison. Potential concerns about test-set label quality or near-duplicate leakage would be threats to external validity or correctness, not circularity, because the claims are measured rather than derived.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the quality of the public dataset, the fairness of the SVM baseline, and the choice of ResNet50 hyperparameters. No new entities are introduced. The free parameters are all standard machine learning hyperparameters, most hand-tuned or left at defaults.

free parameters (9)
  • ResNet50 learning rate = 0.0001
    Hand-selected during manual hyperparameter tuning to improve validation accuracy (Section II-A).
  • ResNet50 batch size = 32
    Manual hyperparameter choice reported as giving better accuracy (Section II-A).
  • ResNet50 number of epochs = 100
    Selected manually; no early stopping or convergence criterion reported (Section II-A).
  • ResNet50 dropout rate = 0.5
    Chosen to reduce overfitting; value not derived from independent evidence (Section II-A).
  • ResNet50 L2 regularization = 0.01
    Chosen for the 1,024-neuron dense layer; tuned on validation (Section II-A).
  • Number of unfrozen ResNet50 layers = 4
    Last four layers unfrozen for fine-tuning; arbitrary modeling choice (Section II-A).
  • Data augmentation strength = 0.2 for rotation, shear, height shift, zoom; horizontal flip
    Manual augmentation configuration; no ablation supports these values (Section II-A).
  • SVM regularization C = 1.0
    scikit-learn default, not tuned; still a free hyperparameter affecting results (Section II-B).
  • SVM gamma = scale
    scikit-learn default; not tuned (Section II-B).
assumptions (4)
  • domain assumption The public African Foods dataset labels are correct and the supplied train/test split is independent and identically distributed.
    All reported accuracies assume that the 251 test images are a valid sample of the target food classes (Section I-A).
  • domain assumption Image-level data augmentation does not bridge the training and test sets.
    Augmentation is applied only to training data; if augmented versions leak into test, accuracy is inflated (Section II-A).
  • domain assumption ImageNet-pretrained ResNet50 features transfer to African food images.
    The fine-tuning strategy assumes that pretrained weights provide a useful starting point for these food categories (Section II-A).
  • standard math Macro and weighted F1 are appropriate summary metrics for an imbalanced six-class problem.
    The paper uses standard definitions of accuracy, precision, recall, and F1 as given in Table II.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Explorative Analysis of SVM Classifier and ResNet50 Architecture on African Food Classification." pith.science (2026). https://pith.science/paper/MND5FKMF

@misc{pith2026250513923,
  author       = {Pith},
  title        = {Pith review of: An Explorative Analysis of SVM Classifier and ResNet50 Architecture on African Food Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MND5FKMF}},
  note         = {Machine review of arXiv:2505.13923}
}
read the original abstract

Food recognition systems has advanced significantly for Western cuisines, yet its application to African foods remains underexplored. This study addresses this gap by evaluating both deep learning and traditional machine learning methods for African food classification. We compared the performance of a fine-tuned ResNet50 model with a Support Vector Machine (SVM) classifier. The dataset comprises 1,658 images across six selected food categories that are known in Africa. To assess model effectiveness, we utilize five key evaluation metrics: Confusion matrix, F1-score, accuracy, recall and precision. Our findings offer valuable insights into the strengths and limitations of both approaches, contributing to the advancement of food recognition for African cuisines.

Figures

Figures reproduced from arXiv: 2505.13923 by the authors.

Figure 1
Figure 1. Sample visualization of Each Class in our Dataset [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Methodology for Fine-Tuning ResNet50 model on African food dataset [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. ResNet50 Architecture [28] During the fine-tuning stage, we used 5-fold cross￾validation, where the model was trained separately on each fold. We manually selected and applied different hyperparam￾eter tuning in a bid to get better results. After training, we saved all the models and selected the one with the highest validation accuracy as the best-performing model. We selected [PITH_FULL_IMAGE:figures/full_fig_p00… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Methodology overview for African food classification using SVM [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 7
Figure 7. Figure 7: Confusion Matrix evaluation report on fine-tuned ResNet50 Architec [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 6
Figure 6. Figure 6: Average Training vs Validation Loss Across 5 Folds in Fine-tuned [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 9
Figure 9. Figure 9: Comparative analysis of performance metrics for ResNet50 and SVM [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [1]

    Automatic Food Recognition Using Deep Convolutional Neural Networks with Self-attention Mechanism

    Abiyev, R., Adepoju, J.(2024). Automatic Food Recognition Using Deep Convolutional Neural Networks with Self-attention Mechanism. Hum- Cent Intell Syst 4, 171–186. https://doi.org/10.1007/s44230-023-00057-9

  2. [2]

    Chaitanya, A., Shetty, J., and Chiplunkar, P. (2023). Food image classi- fication and data extraction using convolutional neural network and web crawlers. Procedia Computer Science, 218, 143-152

  3. [3]

    (2021, March)

    Yadav, S., and Chand, S. (2021, March). Automated food image clas- sification using deep learning approach. In 2021 7th international con- ference on advanced computing and communication systems (ICACCS) (V ol. 1, pp. 542-545). IEEE

  4. [4]

    and Wang, S

    Zhang, Y ., Deng, L., Zhu, H., Wang, W., Ren, Z., Zhou, Q., ... and Wang, S. (2023). Deep learning in food category recognition. Information Fusion, 98, 101859

  5. [5]

    M.(2023)

    Sujay D. M.(2023). Efficient Deep Learning Approach for Food Image Classification. The International journal of analytical and experimental modal analysis. V olume XV , Issue V

  6. [6]

    Usha S.S., Thummaala B.R., Panditi A., and Yerramiya S. Y . A(2022). Automatic Food Image Classification Using Deep Learning Approach. Internation Research journal of Education and Technology. V olume: 04 Issue: 04

  7. [7]

    J., Ejike, C

    Adene, G., Makuo, N. J., Ejike, C. O., Emeka, I. O., and Mbonu, C. E. (2024). Detection and Classification of Human Gender into Binary (Male and Female) Using Convolutional Neural Network (CNN) Model. Asian Journal of Research in Computer Science, 17(6), 135-144

  8. [8]

    H., Das, R., Thepade, S

    Dewan, J. H., Das, R., Thepade, S. D., Jadhav, H., Narsale, N., Mhasawade, A., and Nambiar, S. (2023, April). Image classification by transfer learning using pre-trained CNN models. In 2023 International Conference on Recent Advances in Electrical, Electronics, Ubiquitous Communication, and Computational Intelligence (RAEEUCCI) (pp. 1- 6). IEEE

Show all 32 references
  1. [9]

    Chun, M., Jeong, H., Lee, H., Yoo, T., and Jung, H. (2022). Development of korean food image classification model using public food image dataset and deep learning methods. IEEE Access, 10, 128732-128741

  2. [10]

    (2020, March)

    Phiphiphatphaisit, S., and Surinta, O. (2020, March). Food image classi- fication with improved MobileNet architecture and data augmentation. In Proceedings of the 3rd International Conference on Information Science and Systems (pp. 51-56)

  3. [11]

    M.(2021).Food Classifi- cation Using Deep Learning

    Sridevi G M., Raksha Raj K., and Roshini B. M.(2021).Food Classifi- cation Using Deep Learning. International Journal of Advance Research and Innovative Ideas in Education. V ol-7 Issue-4

  4. [12]

    Xu, B., He, X., and Qu, Z. (2021). Asian food image classification based on deep learning. Journal of Computer and Communications, 9(03), 10

  5. [13]

    E., and Rajaseela, S

    Nithiyaraj, E. E., and Rajaseela, S. (2021). Indian Food Image Recog- nition using a Deep Learning Approach. Indian Journal of Food Engi- neering (IJFE), 1(1)

  6. [14]

    Ogunyale, K. (2020). Convolutional neural network on Nigerian foods

  7. [15]

    A., and Vincent, O

    Iheanacho, C. A., and Vincent, O. R. (2022, November). Classification and recommendation of food intake in West Africa for healthy diet using Deep Learning. In 2022 5th Information Technology for Education and Development (ITED) (pp. 1-6). IEEE

  8. [16]

    Mezgec, S., and Korou ˇsi´c Seljak, B. (2017). NutriNet: a deep learn- ing food and drink image recognition system for dietary assessment. Nutrients, 9(7), 657

  9. [17]

    (2009, November)

    Joutou, T., and Yanai, K. (2009, November). A food image recognition system with multiple kernel learning. In 2009 16th IEEE International Conference on Image Processing (ICIP) (pp. 285-288). IEEE

  10. [18]

    (2007, October)

    Bosch, A., Zisserman, A., and Munoz, X. (2007, October). Image classification using random forests and ferns. In 2007 IEEE 11th international conference on computer vision (pp. 1-8). Ieee

  11. [19]

    (2012, July)

    Matsuda, Y ., Hoashi, H., and Yanai, K. (2012, July). Recognition of multiple-food images by detecting candidate regions. In 2012 IEEE international conference on multimedia and expo (pp. 25-30). IEEE

  12. [20]

    (2010, June)

    Yang, S., Chen, M., Pomerleau, D., and Sukthankar, R. (2010, June). Food recognition using statistics of pairwise local features. In 2010 IEEE computer society conference on computer vision and pattern recognition (pp. 2249-2256). IEEE

  13. [21]

    A., Wanji, S., and Tcheuntue, F

    Abia, W. A., Wanji, S., and Tcheuntue, F. (2007). Energy and nutrient contents of waterfufu and eru. African Journal of Food Science, 1(2), 016-019

  14. [22]

    Ataguba, G., Ezekiel, R., Daniel, J., Ogbuju, E., and Orji, R. (2024). African foods for deep learning-based food recognition systems dataset. Data in Brief, 53, 110092

  15. [23]

    Mandal, B., Okeukwu, A., and Theis, Y . (2021). Masked face recognition using resnet-50. arXiv preprint arXiv:2104.08997

  16. [25]

    H., and Mbonu, C

    Amangeldi, A., Taigonyrov, A., Jawad, M. H., and Mbonu, C. E. (2025). CNN and ViT Efficiency Study on Tiny ImageNet and DermaMNIST Datasets. arXiv preprint arXiv:2505.08259

  17. [26]

    A., and Adebanjo, A

    Al-Haija, Q. A., and Adebanjo, A. (2020, September). Breast cancer diagnosis in histopathological images using ResNet-50 convolutional neural network. In 2020 IEEE International IOT, Electronics and Mecha- tronics Conference (IEMTRONICS) (pp. 1-7). IEEE

  18. [27]

    S., Jim, J

    Islam, T., Hafiz, M. S., Jim, J. R., Kabir, M. M., and Mridha, M. F. (2024). A systematic review of deep learning data augmentation in medical imaging: Recent advances and future research directions. Healthcare Analytics, 100340

  19. [28]

    and Meenakshi Sundaram (2024)

    Anand, M. and Meenakshi Sundaram (2024). Hybrid deep transfer learning and feature fusion architecture for diabetic retinopathy clas- sification and severity grading. Journal of Engineering Science.DOI: 10.52783/jes.4944

  20. [29]

    Muralidharan, V ., Sugumaran, V ., and Sakthivel, N. R. (2011). Wavelet decomposition and support vector machine for fault diagnosis of monoblock centrifugal pump. International Journal of Data Analysis Techniques and Strategies, 3(2), 159-177

  21. [30]

    Cortes, C., and Vapnik, V . (1995). Support-vector networks. Machine learning, 20, 273-297

  22. [31]

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

  23. [32]

    Behar, N., and Shrivastava, M. (2022). ResNet50-Based Effective Model for Breast Cancer Classification Using Histopathology Images. CMES- Computer Modeling in Engineering and Sciences, 130(2)

  24. [33]

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

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.