reference :
http://dcase.community/challenge2017/task-rare-sound-event-detection
The statistics of the duration of the isolated events in seconds are the following:
- babycry, max: 4.82, min: 0.48, mean: 1.80, std: 0.99
- glassbreak, max: 2.16, min: 0.24, mean: 0.85, std: 0.41
- gunshot, max: 2.28, min: 0.20, mean: 0.83, std: 0.58
The original freesound recordings are located at `source_data/events/[classname]/[fileid].wav` and are accompanied by a text file [`[fileid].yaml`] in the same location, which consists of the annotations of the isolated events (field `valid_segments` with start and end times in seconds), as well as other meta data about the recording, as provided on freesound, including the license.
Basically, we have the source sounds file, and the final evaluation dataset is created using mixture code : https://zenodo.org/record/1160455
you could look at the code for the mixture structure.
I hope this helps.
Dear Tariq Sm
Have a look:
https://github.com/ankitshah009/Task-4-Large-scale-weakly-supervised-sound-event-detection-for-smart-cars
Kind Regards
Qamar Ul Islam
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training