search

nuclearboy95 / Anomaly-Detection-PatchSVDD-PyTorch

246 stars 56 forks source link

AUC unusally high during 1st few epochs #11

Open bmabir17 opened 3 years ago

bmabir17 commented 3 years ago

I am training custom object class using my own dataset converted into MvTec Ad format. I am using small lambda value in the results below but its still the same for large lambda values. Screenshot from 2021-03-17 11-58-40 Epochs: 300 Embedding dimension: 64

Can anyone tell me why this is happening? I have also trained 'screw` class and the AUC is high only on 1st epoch and then it goes down drastically.

MuchmorePatience commented 3 years ago

Hi,I'm facing the problem that there is no 'rvai.types' moudle in mvtecad.py,Could u give me a hand?

from rvai.types import String, Image

Traceback (most recent call last): File "main_train.py", line 3, in from codes import mvtecad File "C:\Users\CCone\Desktop\experiment0506\Patch_SVDD-master\codes\mvtecad.py", line 7, in from rvai.types import String, Image ModuleNotFoundError: No module named 'rvai'

bmabir17 commented 3 years ago

@MuchmorePatience this is not related to this issue and i have not faced it. Maybe its a because my development environment is linux. Please check if the package requirements are met properly.

MuchmorePatience commented 3 years ago

@bmabir17 Anyway thanks a lot !

Jiejiegary commented 3 years ago

Hello, guys. I face the similar problem. It seems lamda matters. Can anyone provide optimal lamda for each class?

MuchmorePatience commented 3 years ago

@bmabir17 Hi ,I'm coming again. I failed to understand the 'ground_truth' pictures for what ? Waiting for your reply!