How ishanrai05/skin-cancer-prediction works
This is a PyTorch workflow for classifying HAM10000 dermatoscopic images into seven lesion categories. The principal path is dataset preparation and lesion-aware splitting, image normalization and loading, ResNeXt training and validation, then validation-set evaluation and visual reporting. The README also describes a notebook-based way to run the code and sample-image and class-distribution views. The graph follows relationships visible in the excerpts or README; those views are included as documented capabilities, though their runtime wiring is not established by the excerpts.
Main entry points: main.py Training CLI · train.py Train and validate · utils.py Image dataset · read_data.py Metadata and split
Components 14
Data preparation
- Metadata and split
read_data.py - Image statistics
utils.py - Image dataset
utils.py
Model training
- ResNeXt classifier
main.py - Train and validate
train.py - Training meters
utils.py
Evaluation and views
- Validation evaluation
main.py - Confusion matrix
visualize.py - Sample images
visualize.py - Class distribution
visualize.py
Run interfaces
- Training CLI
main.py - Notebook workflow
notebook/skin-cancer-prediction.ipynb
Other components
- Researcher
- HAM10000 files
Connections 17
- Researcher → Training CLI: starts training
README.md - Researcher → Notebook workflow: runs notebook
README.md - HAM10000 files → Metadata and split: provides metadata
read_data.py - HAM10000 files → Image statistics: provides images
- Training CLI → Metadata and split: requests split
main.py - Training CLI → Image statistics: computes statistics
main.py - Training CLI → Image dataset: constructs datasets
main.py - Image dataset → HAM10000 files: loads image files
utils.py - Training CLI → ResNeXt classifier: configures classifier
main.py - Training CLI → Train and validate: calls train and validate
main.py - Train and validate → ResNeXt classifier: runs forward passes
train.py - Train and validate → Training meters: updates meters
train.py - Training CLI → Validation evaluation: evaluates validation set
main.py - Validation evaluation → ResNeXt classifier: predicts labels
main.py - Validation evaluation → Confusion matrix: plots confusion matrix
main.py - Researcher → Sample images: requests samples
README.md - Researcher → Class distribution: requests counts
README.md