Fix IndexError in sortedAP() when image_average=True in Evaluator

[Frieya]

This pull request fixes an IndexError that occurs when calling e_done.sortedAP() with the image_average parameter set to True. The error is triggered during the printing of the metric in the decorated function due to attempting to index a scalar variable after applying the np.mean operation.

Problem:: When image_average=True, the code attempts to average all metrics, including sortedAP. However, according to the paper, I believe that sortedAP should not be averaged. In this mode, the metric is calculated as the mean of individual metrics from each example. After applying np.mean, the result is a scalar value. The code then tries to index this scalar when processing the sortedAP metric, leading to the following error:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Cell In[17], line 2
      1 # aps contain pairs of (jac, ap), sorted in jac(IoU) increasing order
----> 2 scores, aps = e.sortedAP()
      4 # plot the complete AP curve
      5 aps = np.array(aps)

File G:\Documents\Github\instance_segmentation\evaluate_sortedAP.py:378, in evaluator_decorator.<locals>.decorator.<locals>.decorated(*args, **kwargs)
    376 if verbose:
    377     if metric_name in ['sortedAP']:
--> 378         print('{}: {}, image average: {}'.format(metric_name, metric[0], args[0].image_average))
    379     else:
    380         print('{}: {}, image average: {}'.format(metric_name, metric, args[0].image_average))

IndexError: invalid index to scalar variable.

Solution:

• Modification in the Decorator:
  1. Added sortedAP to the list of metrics that should never be averaged, ensuring it is always treated correctly.