Thank you.
The current version is only a zero-shot (no training) POC for the task of anomaly detection using a pre-trained model.
We are currently working on a few-shot version in order to also see the performance over few-shot optimization (however we are currently running in to some technical issues while trying to fine-tune the pretrained model).
Thank you. The current version is only a zero-shot (no training) POC for the task of anomaly detection using a pre-trained model. We are currently working on a few-shot version in order to also see the performance over few-shot optimization (however we are currently running in to some technical issues while trying to fine-tune the pretrained model).