Keyword Spotting On Google Speech Commands

Metrics

Google Speech Commands V2 35

Results

Performance results of various models on this benchmark

Comparison Table

Model Name	Google Speech Commands V2 35
hts-at-a-hierarchical-token-semantic-audio	98.0
broadcasted-residual-learning-for-efficient	-
work-in-progress-linear-transformers-for	99.1
importantaug-a-data-augmentation-agent-for	95
learning-efficient-representations-for-3	97.0
hello-edge-keyword-spotting-on	-
effective-combination-of-densenet-andbilstm	-
hello-edge-keyword-spotting-on	-
a-neural-attention-model-for-speech-command	93.9
matchboxnet-1d-time-channel-separable-1	-
efficient-keyword-spotting-using-time-delay	-
end-to-end-keyword-spotting-using-neural	-
multi-layer-attention-mechanism-for-speech	-
hello-edge-keyword-spotting-on	-
hello-edge-keyword-spotting-on	-
ast-audio-spectrogram-transformer	98.11
towards-on-device-keyword-spotting-using-low	98.60
temporal-convolution-for-real-time-keyword	-
ssamba-self-supervised-audio-representation	97.4
keyword-transformer-a-self-attention-model	96.95±0.14
attention-free-keyword-spotting	97.56
end-to-end-audio-strikes-back-boosting	98.15
wav2kws-transfer-learning-from-speech	-
training-keyword-spotters-with-limited-and	-
streaming-keyword-spotting-on-mobile-devices	-
micronets-neural-network-architectures-for	-
subspectral-normalization-for-neural-audio	-
hello-edge-keyword-spotting-on	-
pate-aae-incorporating-adversarial	-
keyword-transformer-a-self-attention-model	97.74 ±0.03
keyword-transformer-a-self-attention-model	97.69 ±0.09
training-keyword-spotters-with-limited-and	-
decentralizing-feature-extraction-with	-
subspectral-normalization-for-neural-audio	-
subspectral-normalization-for-neural-audio	-
neural-architecture-search-for-keyword	-
hello-edge-keyword-spotting-on	-
edgecrnn-an-edgecomputing-oriented-model-of	-
convmixer-feature-interactive-convolution	-
howl-a-deployed-open-source-wake-word	-
masked-modeling-duo-learning-representations	98.5
neural-architecture-search-for-keyword	-