Skip to content

GitLab

  • Menu
Projects Groups Snippets
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Sign in
  • bob bob
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
  • Issues 2
    • Issues 2
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 0
    • Merge requests 0
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Environments
    • Releases
  • Monitor
    • Monitor
    • Incidents
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Wiki
    • Wiki
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • bob
  • bobbob
  • Issues
  • #60

Closed
Open
Created Jul 03, 2012 by André Anjos@andre.anjos💬Owner

bob.ip.get_angle_to_horizontal is buggy

Created by: siebenkopf

The C++ implementation of the bob.ip.get_angle_to_horizontal is wrong. To compute the angle between the eyes and the horizontal position, the std::tan function is used instead of the std::atan2. Interestingly, the effect is small when the eye positions are roughly horizontal before, but angles above 20 degrees are calculated wrongly.

The current position of the C++ implementation is located in the file cxx/ip/ip/rotate.h, but used in cxx/ip/ip/FaceEyesNorm.h. Maybe we should move the implementation. Also, we could maybe remove the python bindings of this function (I think, no-one is really using it).

Additionally, there is no unit-test for this function. Also, the bob.ip.FaceEyesNorm, which uses the above mentioned function, is not tested (there is some existing old test code, but it is not executed).

Let's hope that fixing that bug does not influence the face verification results much...

Assignee
Assign to
Time tracking