jayliu0313 / Shape-Guided

Shape-Guided Dual-Memory Learning for 3D Anomaly Detection [ICML2023]
35 stars 6 forks source link

The experimental results differ from official reports #4

Closed EWAN9709 closed 1 year ago

EWAN9709 commented 1 year ago

The experimental hardware configuration is as follows: image

The experimental results are as follows: image

May I ask if our experimental results are normal? Looking forward to your reply

jayliu0313 commented 1 year ago

I think that your RGB scores are not quite right because we use pretrained encoder to get the score, and it would not change the score. In addition, there's a difference in SDF scores compared to ours. May I ask if you trained the SDF model yourself? If not, you should verify that "preprocessing.py" and "cut_patches.py" are executing correctly.