The evaluation table given below consists of the following fields:-

  • Ranking - Algorithms are ranked based on their average joint errors.
  • Algorithm - The name of the algorithm is provided by the user while uploading the result file. Please click on the name of the algorithm to view specific graphical details of its evaluation results.
  • Average Joint Error - This field consists of the average error for all the joints (except for the three thumb-related joints) as measured by our system. It is measured by comparing the data in the uploaded result file with our ground truth data for all the poses and all the test subjects. The average of all the joint errors is then taken and displayed in this field.
  • Median Joint Error - Our system finds the median joint error value from the list of error values obtained for all the poses and all the test subjects, and displays it in this field.
  • Minimum Joint Error - The minimum joint error amongst all the joint error values from all the poses and all the test subjects is displayed in this field.
  • Maximum Joint Error - The maximum joint error amongst all the joint error values from all the poses and all the test subjects is displayed in this field.