Skin disease classification using two path deep transfer learning models
Bibliographic record
Abstract
Skin diseases are among the most common diseases that affect millions of lives per year, yet diagnosing these has several complexities even for trained dermatologists due to overlapping symptoms and features in several diseases. A myriad of deep learning models have been proposed as a solution for diagnosing but a clinically useful model with high accuracy multi disease classification and lower computational complexity is still unavailable. This study focuses on comparing different image pre-processing techniques, transfer learning models and ensemble learning techniques to build a computationally cheap model for 8-class identification of skin diseases. A two path model with EfficientNet and MobileNetV2 transfer learning models as base feature extractors and a final model that stacks the two model results and classifies the images into one of the eight classes is used. The model is trained and tested on ISIC-2019 dataset for 8 class image classification that involves the three types of skin cancers as well. The dataset has an extreme class imbalance problem which leads to favored prediction of the classes with more image, for this first we run simple image augmentation. Secondly, two distinctly processed images are created from each initial image. The two path model takes the two images, gives each to a base model and combines the two outputs, enabling the classifier to consider different features that become prominent due to dissimilar preprocessing techniques. The model is tested with new images on multiple standard metrics to get a final overview of its performance, it gives a diagnosing accuracy of 70% which is close to some state of the art models that consume higher computational power. The classification results imply that by further improving the data gathering and preprocessing techniques along with exploring other base transfer learning models the results of the final model can be reliable while maintaining a low computational requirement, making the diagnoses accessible. This also highlights that such two path algorithms that employ simpler models could be useful for multi class classification tasks where differently processed images might be required to extract features of distinct diseases.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".