How saireddy957/skincancer_project works
This is a research workflow for HAM10000 skin-lesion classification, not a deployed diagnostic service. A user selects training, evaluation, or ACO optimization through the CLI; data loading and preprocessing feed a CNN model, whose checkpoints can then be evaluated and reported. The README also describes image-only and multimodal comparisons, lesion-level splits, and multitask outputs, but sampled implementation excerpts do not establish all of those details; the graph emphasizes verified runtime behavior and notes this evidence limit.
Main entry points: train.py Training and validation · evaluate.py Model evaluation · main.py CLI modes · dataset.py Dataset and transforms
Components 12
Workflow entry
- CLI modes
main.py
Data preparation
- HAM10000 files
- Dataset and transforms
src/dataset.py
Model learning
- CNN model factory
src/model.py - Training and validation
src/train.py - ACO search
src/aco.py - Training utilities
src/utils.py
Evaluation and results
- Model checkpoints
models - Model evaluation
src/evaluate.py - Evaluation artifacts
results - Reported comparison
results/model_comparison.csv
Other components
- Research user
Connections 15
- Research user → CLI modes: selects mode
main.py - CLI modes → Training and validation: starts training
main.py - CLI modes → Model evaluation: starts evaluation
main.py - CLI modes → ACO search: starts search
main.py - HAM10000 files → Dataset and transforms: provides images and metadata
src/dataset.py - Training and validation → Dataset and transforms: loads samples
src/train.py - Training and validation → CNN model factory: creates and trains
src/train.py - Training and validation → Training utilities: saves and plots
src/train.py - Training and validation → Model checkpoints: writes checkpoints
src/train.py - ACO search → Training and validation: trains candidates
src/aco.py - Model evaluation → Dataset and transforms: loads test samples
src/evaluate.py - Model evaluation → CNN model factory: creates model
src/evaluate.py - Model evaluation → Training utilities: loads checkpoint
src/evaluate.py - Model checkpoints → Model evaluation: supplies weights
src/evaluate.py - Model evaluation → Evaluation artifacts: writes reports and plots
src/evaluate.py