Custom anomaly_maps

[davy-blavette]

I don't really understand all these burrs on the heatmap, compratively with a classification I can generate this....is it possible to generate its own heatmap, currently I use a callback

def on_test_batch_end(self, trainer, pl_module, outputs, batch, batch_idx):
        images = batch[‘image’]
        image_paths = batch[‘image_path’]]
        anomaly_maps = outputs[‘anomaly_maps’]

but I would like access to the anomaly_maps property, or to be able to generate it myself.
Anomalib

My heatmap

[alexriedel1]

I don't really understand all these burrs on the heatmap, compratively with a classification I can generate this.

What do you mean by burrs? Where is the actual anomaly in your image that you want to detect? What do you mean by generating this with a classification?