xiahaifeng1995/PaDiM-Anomaly-Detection-Localization-master

Issues

• saving the trained model

  #46 opened 10 months ago by gurbir739
  0

• Speed up the PaDim

  #45 opened a year ago by DeepKnowledge1
  0

• May I ask if there are any good optimization methods for models that often detect normal samples as anomalies

  #42 opened a year ago by cvcabbage
  0

• how to deploy this method in an industry application

  #31 opened 3 years ago by mrljwlm
  3

• while embedding_concat() runs, I got "Killed"

  #32 opened 3 years ago by wonchul-kim
  5

• 内存溢出

  #28 opened 3 years ago by 102757017
  6

• GMM and Mahalanobis distance

  #41 opened a year ago by Leonardo0325
  0

• Pro-score

  #39 opened 2 years ago by tangyaorui-tyr
  0

• pro-score指标

  #38 opened 2 years ago by tangyaorui-tyr
  0

• PaDiM-EfficientNetV2, highest image-level ROCAUC=0.978

  #37 opened 2 years ago by JohnnyHopp
  3

• ValueError: continuous format is not supported

  #24 opened 3 years ago by AlgirdasKartavicius
  2

• Cuda is slower than cpu

  #17 opened 4 years ago by dhkdnduq
  6

• KILLED problem

  #35 opened 2 years ago by aytek03
  2

• What is the evaluation protocol in paper?

  #36 opened 2 years ago by edwardcho
  0

• Whether the object position is strictly required to be consistent？

  #1 opened 4 years ago by GlassyWing
  4

• Question about CPU out of memory

  #27 opened 3 years ago by Youskrpig
  10

• If I want to make the input image size as 512, what shoud I do?

  #30 opened 3 years ago by wonchul-kim
  1

• about embedding concat

  #11 opened 4 years ago by chengweige
  3

• Calculate image ROCAUC score low

  #34 opened 3 years ago by VivizSun
  0

• Inference time

  #8 opened 4 years ago by sangkyuleeKOR
  25

• when i test，the time is so long

  #33 opened 3 years ago by leolv131
  4

• How do you approach when calculating score_map per batch in inference?

  #29 opened 3 years ago by ingbeeedd
  1

• About the size of the input image

  #9 opened 4 years ago by xfby2016
  3

• Cut the model to save time

  #13 opened 4 years ago by DeepKnowledge1
  5

• PaDim without Ground Truth Labels + Using self-defined dataset

  #25 opened 3 years ago by beaverCraze
  3

• The score is very different depending on the random idx value

  #15 opened 4 years ago by dhkdnduq
  1

• What is criteria of "t_d" and "d" in main code? and efficientnet-b5

  #22 opened 4 years ago by GaengSu
  3

• Can i use another models?

  #23 opened 4 years ago by GaengSu
  2

• Question about more than one anomalies in one image？

  #21 opened 4 years ago by Youskrpig
  1

• LICENSE file

  #10 opened 4 years ago by SamvitJ
  4

• Is it necessary to embed layer1?

  #19 opened 4 years ago by dhkdnduq
  0

• threshold of anomaly classification

  #20 opened 4 years ago by QQR1
  0

• The score is very different

  #18 opened 4 years ago by DeepKnowledge1
  1

• Apply PaDiM to another domain.

  #16 opened 4 years ago by sangkyuleeKOR
  8

• PRO Score

  #14 opened 4 years ago by prakhar123pradhan
  0

• It also looks good to save conv_inv.

  #12 opened 4 years ago by dhkdnduq
  1

• Threshold selection for the testing

  #7 opened 4 years ago by xfby2016
  2

• Results for R18-Rd100?

  #6 opened 4 years ago by TzxzT
  0

• Large data

  #2 opened 4 years ago by DeepKnowledge1
  2

• About the PRO-score

  #5 opened 4 years ago by Classmate-Huang
  2

• Error

  #3 opened 4 years ago by reddevil1310
  3