Efficient Large-scale Audio Tagging via Transformer-to-CNN Knowledge Distillation

Benchmark Model Rank Results
audio-classification-on-esc-50mn40_as#5Top-1 Accuracy: 97.45PRE-TRAINING DATASET: AudioSet…
audio-tagging-on-audiosetmn40_as (Ensemble)#5mean average precision: 0.498
audio-tagging-on-audiosetmn40_as (Single)#10mean average precision: 0.483
audio-classification-on-audiosetmn40_as (Ensemble)#9Test mAP: 0.498
audio-classification-on-audiosetmn40_as (Single)#22Test mAP: 0.483