ROC SDK  2.4.0
Scalable Face Recognition Software
roc-compare

Measure the similarity between galleries.

This command line application corresponds to the roc_compare_galleries function in the Application Programming Interface.

Options

Option Meaning Default
-b Write a Binary Similarity Matrix Write a text similarity matrix
–headers Write row and column headers using the image file base names Don't write row and column headers

Binary Similarity Matrix

By default similarity scores are written in tab-separated plain text. For large similarity matricies it is more efficient to store the scores in binary form. When -b is specified, the output of comparison will be an MBGC-compatible Similarity Matrix.

Examples

Show help
Print application usage and supported arguments.
$ roc-compare -h 
Measure the similarity between two templates
Print the similarity between two templates faceA.t and faceB.t.
$ roc-compare faceA.t faceB.t 
Construct a similarity matrix
Write the pairwise similarity between all templates in facesA.t and facesB.t to similarity.tsv.
$ roc-compare facesA.t facesB.t similarity.tsv 
Construct a binary similarity matrix
Write the pairwise similarity between all templates in facesA.t and facesB.t in binary to similarity.mtx.
$ roc-compare -b facesA.t facesB.t similarity.mtx 
Construct a self-similarity matrix
Write the pairwise similarity between all teamplates in faces.t to self_similarity.tsv with row/column headers.
$ roc-compare --headers faces.t faces.t similarity.tsv 
Construct a similarity matrix to stdout
Print the pairwise similarity between all templates in facesA.t and facesB.t to stdout.
$ roc-compare facesA.t facesB.t > similarity.tsv