Please ask about problems and questions regarding this tutorial on answers.ros.org. Don't forget to include in your question the link to this page, the versions of your OS & ROS, and also add appropriate tags.

Processing Results with the Stair Vision Library

From person_data/evaluation, run this script:

evaluate.py <detection dir> <annotation dir> <output dir> Finds all .xml files in detection dir and annotation dir. Creates output dir and adds the following to it:

• missing_detections.txt: A list of all annotation files that did not have a corresponding detection file missing_annotations.txt: A list of all detection files that did not have a corresponding annotation file svl_detections.xml: The detections in svl format svl_annotations.xml: The bounding box annotations in svl format svl_annotations_tight.xml: The tightest possible bounding boxes surrounding each set of polygons, in svl format pr.person.txt: PR curve in CSV format generated by scoring svl_detections against svl_annotations pr_tight.person.txt: PR curve in CSV format generated by scoring svl_detections against svl_annotations_tight

Frames containing bad images of people that you do not want scored should not have XML files.