Dense Semantic Correspondence where Every Pixel is a Classifier
Hilton Bristow, Jack Valmadre and Simon Lucey,"Dense Semantic Correspondence where Every Pixel is a Classifier",International Conference on Computer Vision (ICCV), 2015
EPiC solves the dense semantic correspondence problem by constructing an LDAclassifier around every pixel in the source image, and convolving it withevery point in the target image to produce a probability likelihood estimate.
The best correspondence is then estimated by regularizing the likelihoodwith spatial constraints.
Usingpip, the repository can be cloned and built automatically:
pip install git+https://github.com/hbristow/epic
The requirements are pure-Python, and will be retrieved automatically
The initial public release of this research only contains code to build andapply detectors on image pairs. It does not contain functionality to performregularization. We are working to provide wrappers to SIFT Flow.