MetricsVisualizer Help

[markstanl]

I am struggling to use the MetricsVisualizer() class, and cannot find any relevant documentation to help me.
I am following what the docstrings say in the file, and have this code

from anomalib.data import MVTecAD
from anomalib.engine import Engine
from anomalib.models import EfficientAd
from anomalib.utils.visualization import MetricsVisualizer

datamodule = MVTecAD(
    root="./datasets/MVTecAD",  # Path to download/store the dataset
    category="bottle",  # MVTec category to use
    train_batch_size=1,  # Number of images per training batch
    eval_batch_size=1,  # Number of images per validation/test batch
    num_workers=8,  # Number of parallel processes for data loading
)
model = EfficientAd()
engine = Engine(max_epochs=2)
engine.fit(datamodule=datamodule, model=model)

visualizer = MetricsVisualizer()
results = visualizer.generate(pl_module=model)
for result in visualizer.generate(pl_module=model):
    print(result.file_name)

Which yields result:

AttributeError: 'EfficientAd' object has no attribute 'image_metrics'

I understand MetricsVisualizer inherits BaseVisualizer, which is deprecated in place of Visualizer, however it seems like there is no new version. Are there plans to re-implement it?

[samet-akcay]

@markstanl, thanks for bringing this to our attention. Yeah we had some plans, but it's been on our backlog.

torchmetrics has plot method that we might need to adapt. I don't have a specific design in my mind currently though.

[samet-akcay]

fyi @alfieroddan @ashwinvaidya17 @rajeshgangireddy

[markstanl]

I was really just trying to visualize the ROC curve.
My current workaround until that is implemented is to use the metrics generate_figure method, which does exactly what I wanted.

from anomalib.data import MVTecAD
from anomalib.engine import Engine
from anomalib.models import EfficientAd
from anomalib.metrics import AUROC

datamodule = MVTecAD(
    root="./datasets/MVTecAD",  # Path to download/store the dataset
    category="bottle",  # MVTec category to use
    train_batch_size=1,  # Number of images per training batch
    eval_batch_size=1,  # Number of images per validation/test batch
    num_workers=8,  # Number of parallel processes for data loading
)
model = EfficientAd()
engine = Engine(max_epochs=2)
engine.fit(datamodule=datamodule, model=model)

res = engine.predict(model=model, dataloaders=datamodule.test_dataloader())
auroc = AUROC(fields=["anomaly_map", "gt_mask"])

for batch in res:
    auroc.update(batch)

fig, _ = auroc.generate_figure()
fig.savefig("roc_curve.png")