This repository, contains examples on how to extract descriptors from sequences of the HPatches dataset.
The HPatches evaluation protocol expects a .csv file per .png
image, with N rows and D columns, where N is the number of
patches in the sequence, and D is the descriptor dimensionality.
For example, the v_dogman\ref.png image is of size 38350x65, thus it
consists of 38350/65=590 patches. It follows that a valid descriptor
.csv file for this image, would consist of 590 rows.
We provide examples for several languages, together with guides on how
to use template to extract the trivial (μ,σ) descriptor from
HPatches.
Please check the individual readme files for language-specific information.
In the folder baselines we collect several example implementations
of computing descriptors for HPatches sequences, for previously
published baseline descriptors.