English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Data-Free Knowledge Distillation (DFKD) has recently attracted growingattention in the academic community, especially with major breakthroughs incomputer vision. Despite promising results, the technique has not been wellapplied to audio and signal processing. Due to the variable duration of audiosignals, it has its own unique way of modeling. In this work, we proposefeature-rich audio model inversion (FRAMI), a data-free knowledge distillationframework for general sound classification tasks. It first generateshigh-quality and feature-rich Mel-spectrograms through a feature-invariantcontrastive loss. Then, the hidden states before and after the statisticspooling layer are reused when knowledge distillation is performed on thesefeature-rich samples. Experimental results on the Urbansound8k, ESC-50, andaudioMNIST datasets demonstrate that FRAMI can generate feature-rich samples.Meanwhile, the accuracy of the student model is further improved by reusing thehidden state and significantly outperforms the baseline method.

Feature-Rich Audio Model Inversion for Data-Free Knowledge Distillation Towards General Sound Classification