Update README.md

Added information on how to use the post-processing script to get predictions based on class-specific count thresholds.

README.md

@@ -53,6 +53,13 @@ The model can be applied on all wav files present in a folder (`<data-root>`) an
 autrainer inference hf:autrainer/edansa-2019-cnn10-32k-t <data-root> <output-root>
 ```
 
+In order to obtain the predictions based on the class-specific count thresholds, we recommend using a window size of 10s and a hop size of 1s (`-w 10 -s 1`).
+Then apply the postprocess_predictions.py script to obtain the final predictions by specifying the csv path (`--path`) to the results.csv:
+
+```
+python postprocess_predictions.py --path /path/to/results.csv
+```
+
 ## Training
 
 ### Pretraining