How do I output results for each category during validation?

[d710055071]

How do I output results for each category during validation?
for example
precision recall f1-score support

       5     0.9272    0.9955    0.9601      1331
      10     0.1672    0.7183    0.2713        71
      15     0.2586    0.8824    0.4000        17
      20     0.8351    0.9766    0.9003      1110
      30     0.9873    0.9455    0.9660      6886
      40     0.9690    0.9767    0.9729      7266
      50     0.9948    0.7495    0.8549      5074
      60     0.8506    0.9758    0.9089      5380
      70     0.9986    0.9257    0.9608      1575
      80     0.9233    0.9666    0.9445      3738
     100     0.9657    0.9876    0.9765      1369
     120     0.2679    0.9375    0.4167        16

accuracy                         0.9332     33833

macro avg 0.7621 0.9198 0.7944 33833
weighted avg 0.9456 0.9332 0.9351 33833

[Ezra-Yu]

in config file:

val_evaluator = dict(type='ConfusionMatrix')

or

val_evaluator = [dict(type='ConfusionMatrix'), ...]

[Xuchangsheng1]

Where is inserted into?

[Ezra-Yu]

in the dataset section.

[whdcumt]

in config file:

val_evaluator = dict(type='ConfusionMatrix')

or

val_evaluator = [dict(type='ConfusionMatrix'), ...]

This method should be set in version 0.x，In version 1.x, this setting will be reported with an error，，In version 1.x，Under the folder /mmcls/evaluation/metrics, there are 'Accuracy', 'SingleLabelMetric', 'MultiLabelMetric', 'AveragePrecision',
'MultiTasksMetric', 'VOCAveragePrecision', 'VOCMultiLabelMetric'
My question is, 'Accuracy', 'SingleLabelMetric',Whether they can be output at the same time,There is no problem with single metric output，Also, is there code for calculating AUC values and plotting VOC curves

[whdcumt]

I got it in the instruction manual

[Ezra-Yu]

there is a PR https://github.com/open-mmlab/mmpretrain/pull/1345/files#diff-3f842b63c42e45106f28b9b3c7069a1a739a747e98c332d158c4a802279f88b6 for AUC

[fansofml]

you can redefine the metric and print it