Anomalib CLI Improvements - Update metrics and create post_processing section in the config file.

[samet-akcay]

The new Anomalib CLI has threshold and normalization_method parameters in the metrics section. However, this may not be intuitive. Instead, it would be better to create a post_processing to store these parameters.

In addition, image_metric_names and pixel_metric_names are a bit verbose, especially for the API version. Instead, those could be renamed as image_metrics and pixel_metrics, respectively.

Overall, the updated config file would be something like:

.
.
.
metrics:
  image_metrics:
    - F1Score
    - AUROC
  pixel_metrics:
    - F1Score
    - AUROC

post_processing:
  adaptive_threshold: true
  default_image_threshold: null
  default_pixel_threshold: null
  normalization_method: min_max

visualization:
  task: segmentation
  log_images_to:
    - local
  inputs_are_normalized: true
.
.
.

• Update the metrics_names parameters
• Assign default values to metrics_names
• Move threshold and normalization_method parameters to post_processing section.

[cyj95]

where are metrics calculated? in which python file