Investigating the Quality of DermaMNIST and Fitzpatrick17k Dermatological Image Datasets
<h2>Abstract</h2>
<p>The remarkable progress of deep learning in dermatological tasks has brought us closer to achieving diagnostic accuracies comparable to those of human experts. However, while large datasets play a crucial role in the development of reliable deep neural network models, the quality of data therein and their correct usage are of paramount importance. Several factors can impact data quality, such as the presence of duplicates, data leakage across train-test partitions, mislabeled images, and the absence of a well-defined test partition. In this paper, we conduct meticulous analyses of three popular dermatological image datasets: DermaMNIST, its source HAM10000, and Fitzpatrick17k, uncovering these data quality issues, measure the effects of these problems on the benchmark results, and propose corrections to the datasets. Besides ensuring the reproducibility of our analysis, by making our analysis pipeline and the accompanying code publicly available, we aim to encourage similar explorations and to facilitate the identification and addressing of potential data quality issues in other large datasets.</p>
<h2>Citation</h2>
<p>If you find this project useful or if you use our newly proposed datasets and/or our analyses, please cite our paper.</p>
<blockquote>
<pre>Kumar Abhishek, Aditi Jain, Ghassan Hamarneh. "Investigating the Quality of DermaMNIST and Fitzpatrick17k Dermatological Image Datasets". arXiv preprint arXiv:2401.14497, 2024. DOI: 10.48550/ARXIV.2401.14497.</pre>
</blockquote>
<p>The corresponding BibTeX entry is:</p>
<blockquote>
<p><code>@article{abhishek2024investigating,</code><br><code> title={Investigating the Quality of {DermaMNIST} and {Fitzpatrick17k} Dermatological Image Datasets},</code><br><code> author={Abhishek, Kumar and Jain, Aditi and Hamarneh, Ghassan},</code><br><code> journal={arXiv preprint arXiv:2401.14497},</code><br><code> doi = {10.48550/ARXIV.2401.14497},</code><br><code> url = {https://arxiv.org/abs/2401.14497},</code><br><code> year={2024}</code><br><code>}</code></p>
</blockquote>
<h2>Project Website</h2>
<p>The results of the analysis, including the visualizations, are available on the project website: <a href="https://derm.cs.sfu.ca/critique/" target="_blank" rel="noopener">https://derm.cs.sfu.ca/critique/</a>.</p>
<h2>Code</h2>
<p>The accompanying code for this project is hosted on GitHub at <a title="Corrected-Skin-Image-Datasets" href="https://github.com/kakumarabhishek/Corrected-Skin-Image-Datasets" target="_blank" rel="noopener">https://github.com/kakumarabhishek/Corrected-Skin-Image-Datasets</a>.</p>
<h2>License</h2>
<p>The metadata files (<code>DermaMNIST-C.csv</code>, <code>DermaMNIST-E.csv</code>, <code>Fitzpatrick17k_DiagnosisMapping.xlsx</code>,<code>Fitzpatrick17k-C.csv</code>) contained in this repository are licensed under <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank" rel="noopener">the Creative Commons Attribution 4.0 International (<strong>CC BY 4.0</strong>) License</a>.</p>
<p>The NPZ files associated with DermaMNIST-C (<code>dermamnist_corrected_28.npz</code>, <code>dermamnist_corrected_224.npz</code>) and DermaMNIST-E (<code>dermamnist_extended_28.npz</code>, <code>dermamnist_extended_224.npz</code>) contained in this repository are licensed under <a href="https://creativecommons.org/licenses/by-nc/4.0/" target="_blank" rel="noopener">the Creative Commons Attribution-NonCommercial 4.0 International (<strong>CC BY-NC 4.0</strong>) License</a>.</p>
<p>The code hosted on <a href="https://github.com/kakumarabhishek/Corrected-Skin-Image-Datasets" target="_blank" rel="noopener">GitHub</a> is licensed under <a href="https://github.com/kakumarabhishek/Corrected-Skin-Image-Datasets/blob/main/LICENSE" target="_blank" rel="noopener">the Apache License 2.0</a>.</p>
opencc-by-4.0Jul 2023View details →
