❓How to use F1Max

[jun614]

Hello, using F1Max() directly may result in an error, but I can see from the example of f1_core.exe that there is no problem
Traceback (most recent call last):
File "/home/sjx/project/LogSAD/evaluation.py", line 164, in
run(args.module_path, args.class_name, args.weights_path, args.dataset_path, args.category, args.viz)
File "/home/sjx/project/LogSAD/evaluation.py", line 92, in run
image_metric = F1Max()
File "/home/sjx/Downloads/anomalib-main/src/anomalib/metrics/base.py", line 132, in init
raise ValueError(msg)
ValueError: Batch fields must be provided for metric <class 'anomalib.metrics.f1_score.F1Max'>. Use the fields argument to specify which fields from the batch object should be used to update the metric.

[rajeshgangireddy]

Hi @jun614
Please see related issue : #2737
Would that solve your issue?

[samet-akcay]

It's hard to tell what is in /home/sjx/project/LogSAD/evaluation.py, but I suspect that you are not using Anomalib's Batch object.

You have two options:

• You could follow @rajeshgangireddy's suggestion in Metric Usage Example get Error #2737
• Alternatively, you could use the _F1Max class that does not need Batch but just uses torch.Tensor. Something like the following:

anomalib/src/anomalib/metrics/f1_score.py

Lines 63 to 97 in f60dd31

	class _F1Max(Metric):
	"""F1Max metric for computing the maximum F1 score.
	This class calculates the maximum F1 score by varying the classification
	threshold. The F1 score is the harmonic mean of precision and recall,
	providing a balanced metric for imbalanced datasets.
	After computing the maximum F1 score, the class stores the threshold that
	achieved this score in the ``threshold`` attribute.
	Args:
	**kwargs: Additional arguments passed to the parent ``Metric`` class.
	Attributes:
	full_state_update (bool): Whether to update entire state on each batch.
	Set to ``False`` as metric only needs current batch.
	precision_recall_curve (BinaryPrecisionRecallCurve): Utility to compute
	precision-recall values across thresholds.
	threshold (torch.Tensor): Threshold value that yields maximum F1 score.
	Example:
	>>> from anomalib.metrics import F1Max
	>>> import torch
	>>> # Create metric
	>>> f1_max = F1Max()
	>>> # Create sample data
	>>> preds = torch.tensor([0.1, 0.4, 0.35, 0.8])
	>>> target = torch.tensor([0, 0, 1, 1])
	>>> # Update and compute
	>>> f1_max.update(preds, target)
	>>> f1_max.compute()
	tensor(1.0)
	>>> f1_max.threshold
	tensor(0.6000)
	"""