List your company
ResearchFormerly WizWand

Environmental Sound Classification on ESC-50 (5-fold cross-validation)

98.1AccuracyBEATS
35.4451.707567.97584.2425Dec 21, 2019Jan 7, 2021Jan 25, 2022Feb 13, 2023Mar 2, 2024Mar 20, 2025Apr 8, 2026
Updated 6mo ago

Evaluation results

MethodLinks
BEATS2023.1098.1
DyMN-LWidth Multiplier (alpha)=2.02023.1097.4
MNWidth Multiplier (alpha)=2.02023.1096.9
PaSST-S2023.1096.8
MNWidth Multiplier (alpha)=3.02023.1096.8
CLAP2023.1096.7
DyMN-SWidth Multiplier (alpha)=0.42023.1096.4
MNWidth Multiplier (alpha)=1.02023.1096.4
DyMN-MWidth Multiplier (alpha)=1.02023.1096.4
AST-PAdditional audio data pre-training=true2021.0495.6
CNN14Training strategy=Fine-tune2019.1294.7
SOTA-PAdditional audio data pre-training=true2021.0494.7
MNWidth Multiplier (alpha)=0.42023.1093.2
BEATs#Param. (M)=91, Data (hrs)=5,800, Evaluation Protocol=Fine-tuning2026.0492.45
CNN14Training strategy=Freeze, Classifier layers=3 fully-connected layers2019.1291.8
ULTRAS#Param. (M)=89, Data (hrs)=2,000, Evaluation Protocol=Fine-tuning2026.0491.15
CNN14Training strategy=Freeze, Classifier layers=1 fully-connected layer2019.1290.8
Cross-AVIDPretraining DB=Audioset, Evaluation protocol=linear classification2020.0489.2
AVID+CMAPretraining DB=Audioset, Evaluation protocol=linear classification2020.0489.1
AST-SAdditional audio data pre-training=false2021.0488.7
EnCodecMAE#Param. (M)=86.6, Data (hrs)=11,000, Evaluation Protocol=Fine-tuning2026.0488.25
Sailor et al.2019.1286.5
ConvRBMPretraining DB=None, Evaluation protocol=linear classification2020.0486.5
SOTA-SAdditional audio data pre-training=false2021.0486.5
XDCPretraining DB=Audioset, Evaluation protocol=linear classification2020.0485.8
CNN14Training strategy=Scratch2019.1283.3
AVTSPretraining DB=Audioset, Evaluation protocol=linear classification2020.0480.6
L3Pretraining DB=Flickr-SoundNet, Evaluation protocol=linear classification2020.0479.3
AVID+CMAPretraining DB=Kinetics, Evaluation protocol=linear classification2020.0479.1
XDCPretraining DB=Kinetics, Evaluation protocol=linear classification2020.0478.5
Cross-AVIDPretraining DB=Kinetics, Evaluation protocol=linear classification2020.0477.6
HuBERT#Param. (M)=95, Data (hrs)=960, Evaluation Protocol=Fine-tuning2026.0477.45
ESC50 Baseline2023.1076.9
AVTSPretraining DB=Kinetics, Evaluation protocol=linear classification2020.0476.7
SoundNetPretraining DB=Flickr-SoundNet, Evaluation protocol=linear classification2020.0474.2
ConvNetPretraining DB=None, Evaluation protocol=linear classification2020.0464.5
RandomForestPretraining DB=None, Evaluation protocol=linear classification2020.0444.3
SSAST#Param. (M)=89, Data (hrs)=5,800, Evaluation Protocol=Fine-tuning2026.0437.85