A framework for pattern recognition pipelines.
This framework here is one of my "bigger" spare-time projects. It's intended to get developers up and running with a pattern recognition pipeline.
It evolved from my work on my master's project at the university. While working on my project, I needed to build a pipeline to perform text detection on historical document images.
It should include the following steps:
- Preprocessing
- Feature Extraction
- Dimensionality Reduction
- Encoding
- Classification
Since doing things on your own actually helps you to learn lots of things, I started from scratch. But along my way I soon realized that a lot of the boilerplate I was writing to glue my pipeline together would also be written by the next student who would need a PR pipeline.
Since I like to design frameworks, I decided to build a modular framework for pattern recognition pipelines in my spare-time.
-
Utils
- Logging
- I/O
- Commandline Arguments
- k-fold crossvalidation
- Platt scaling
-
Core
- Clustering
- K-Means
- DBSCAN
- GMM
- Dimensionality Reduction
- PCA
- Encoding
- BoW
- VLAD
- Fisher Vectors
- Feature Extraction
- SIFT
- Zernike Moments
- OTC
- Machine Learning
- Linear SVM
- Preprocessing
- Vesselfilter
- Binarization
- Postprocessing
- Density Based Filtering
- Clustering
-
GUI
- User Interface for Pipeline Tools
This list is by no means complete and will be extended in the future.
This project uses awesome open source libraries, so thank you very much for your work!