How mickramygoseng/cnn-vs-svm-based-baybayin-ocr works
This is a Baybayin OCR research pipeline: image samples are fetched and converted into SVM features and CNN pixels, split into repeated stratified folds, then evaluated by both models with metrics saved for comparison. Dataset sanity and cross-model correspondence checks are a separate supporting workflow. The sampled source confirms these stages and their file-based handoffs; MANOVA is mentioned in the README but has no implementation represented in the supplied tree.
Main entry points: evaluate.py Evaluation runner · preprocess.py Preprocessing runner · dataset_validation.py Sanity validator · sample.py Sampling runner
Components 15
Image preparation
- Raw samples
- Image fetcher
modules/pre_processing/get_image_paths.py - Preprocessing runner
modules/pre_processing/preprocess.py - Image conversion
modules/pre_processing/preprocess_images.py - Processed datasets
Fold sampling
- Sampling runner
modules/sampling/sample.py - Stratified splitter
modules/sampling/sampling.py - Sampled folds
Model evaluation
- Evaluation runner
modules/evaluation/evaluate.py - SVM model
modules/evaluation/models/SVM.py - CNN model
modules/evaluation/models/CNN.py - Metrics results
Dataset validation
- Sanity validator
dataset_validation.py - Validation checks
Other components
- Researcher
Connections 21
- Researcher → Raw samples: provides samples
README.md - Researcher → Preprocessing runner: runs preprocessing
README.md - Preprocessing runner → Image fetcher: fetches paths
modules/pre_processing/preprocess.py - Image fetcher → Raw samples: reads images
modules/pre_processing/get_image_paths.py - Preprocessing runner → Image conversion: converts images
modules/pre_processing/preprocess.py - Preprocessing runner → Processed datasets: writes datasets
modules/pre_processing/preprocess.py - Researcher → Sampling runner: runs sampling
README.md - Sampling runner → Processed datasets: loads datasets
modules/sampling/sample.py - Sampling runner → Stratified splitter: runs splits
modules/sampling/sample.py - Stratified splitter → Sampled folds: writes folds
modules/sampling/sampling.py - Researcher → Evaluation runner: runs evaluation
README.md - Evaluation runner → Sampled folds: loads fold data
modules/evaluation/evaluate.py - Evaluation runner → SVM model: runs SVM
modules/evaluation/evaluate.py - Evaluation runner → CNN model: runs CNN
modules/evaluation/evaluate.py - SVM model → Evaluation runner: returns metrics
modules/evaluation/models/SVM.py - CNN model → Evaluation runner: returns metrics
modules/evaluation/models/CNN.py - Evaluation runner → Metrics results: saves metrics
modules/evaluation/evaluate.py - Researcher → Sanity validator: runs checks
README.md - Sanity validator → Processed datasets: checks datasets
dataset_validation.py - Sanity validator → Sampled folds: checks fold data
dataset_validation.py - Sanity validator → Validation checks: prints findings
dataset_validation.py