Training 100 epochs, AUC only 51

[lqs378140341]

Only the development dataset of DCASE 2020 Task 2 is used，include fan、pump、slider、Toycar、toyconveyor、value data，what should the class_num be set to?
Thanks for answering~

[liuyoude]

The class_num is set as the number of IDs in train dataset, and if only use development dataset the code related to getting the training labels also needs to be modified.

[lqs378140341]

ToyCar id,AUC,pAUC 12,0.7614880952380951,0.6448507632718159 13,0.9269595687331535,0.8361753440204285 14,0.9902964959568733,0.952844375088665 15,0.9983611859838276,0.9913746630727762 Average,0.9192763364779873,0.8563112863634214 ToyConveyor id,AUC,pAUC 16,0.8079750000000001,0.6695394736842105 17,0.6025228873239437,0.5203854707190512 18,0.7079536039768018,0.5987627232335453 Average,0.7061504971002486,0.5962292225456024 fan id,AUC,pAUC 00,0.44742014742014746,0.5237294710978921 01,0.7056267409470752,0.5348189415041783 02,0.5996232269503546,0.5212765957446809 03,0.8931163434903047,0.8547893278903631 Average,0.6614466147019705,0.6086535840592786 pump id,AUC,pAUC 04,0.7319023569023569,0.6271929824561404 05,0.7506306306306306,0.5656709340919868 06,0.48707482993197276,0.5333572025301349 07,0.7251960784313726,0.500515995872033 Average,0.6737009739740832,0.5566842787375738 slider id,AUC,pAUC 08,0.9992977528089888,0.9965996451803667 09,0.9070411985018726,0.6944608712793219 10,0.9997752808988764,0.9988172678888232 11,0.9765168539325844,0.8775872264931993 Average,0.9706577715355805,0.8918662527104277 valve id,AUC,pAUC 19,1.0,1.0 20,0.6754583333333333,0.5859649122807018 21,0.9857499999999999,0.9640350877192982 22,0.9476666666666667,0.7969298245614036 Average,0.90221875,0.8367324561403509 Total Average,0.8055751572983118,0.7244128467594425

The difference between the recognition results of only changed categories and your training results is too large. Is there anything wrong with it？