PSLA: Improving Audio Tagging with Pretraining, Sampling, Labeling, and Aggregation

Benchmark Model Rank Results
audio-classification-on-fsd50kPSLA#7mAP: 56.71
audio-tagging-on-audiosetPSLA#11mean average precision: 0.474
audio-classification-on-audiosetPSLA (Ensemble)#28Test mAP: 0.474AUC: 0.981d-prime: 2.936
audio-classification-on-audiosetPSLA (Single)#36Test mAP: 0.443AUC: 0.975d-prime: 2.778