Training corpus summary
projection of the training data. Each dot is one window of audio (explanation below). Hover a dot to hear the sound; click to open the recorded sample.
| Label | Duration | Occupancy | Samples | Fragments | Fragment durations0.5s bins · hover for counts | Windows | Window durations0.1s bins · hover for counts |
|---|---|---|---|---|---|---|---|
| bark | 1–3 s | 50–80 % | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| yap | 1–3 s | 50–80 % | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| background | 3–5 s | n/a | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| wind | 3–5 s | n/a | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| homestead | 3–5 s | n/a | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| traffic | 3–5 s | n/a | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| wildlife | 1–3 s | 50–80 % | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| gunshot | 1–3 s | 50–80 % | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
| wrongdog | 1–3 s | 50–80 % | 0 | 0 | {b:0,0,0,0,0,0,0,0,0,0,0} | — | {b:0,0,0,0,0,0,0,0,0,0,0} |
Desired number of fragments: <30 not enough · 30–99 viable · 100+ good
Hover the table row above to highlight the dots in scatterplot. Click to select.
Each dot is a window. What are YAMNet windows and how are they made?
