This task requires the query by using exact but noisy recordings.
- 589 songs (*.mp3 and *.wav) corresponding to 573 different songs
- Language: English, Chinese, Japanese, ...
- 35 mono, 554 stereo
- 40.97 hours long
- 305 recordings (*.wav) corresponding to 304 different songs in DB
- 166 mono, 139 stereo
- 16.11 hours long
- Recorded by variety brand of smartphones, containing noise
- Variety of sampling rate and bit resolution
- Divided into 5800 10-second clips
Top-10 hit rate (1 point is scored for a hit in the top 10 and 0 is scored otherwise, as in the Query by Singing/Humming task)
Participants are required to submit a breakdown version of algorithm. The two parts are:
- Fingerprinting extractor
where %file.list% is the input list of database audio files named as uniq_key.mp3. or uniq_key.wav For example:
extractor %file.list% %dir % %database_or_query% ./AFP/database/00001.wav
Time and hardware limits
Due to the potentially high number of participants in this and other audio tasks, hard limits on the runtime of submissions are specified. The time/storage limits of different steps are shown in the following table:
|Steps||Time limit||Storage (hard disk) limit|
|extractor (for database)||Totally 72 hours||40KB for every 10 second audio + 250 MB extra overhead (if needed)|
|extractor (for query)||40 KB for each query clip|
Submissions that exceed these limitations may not receive a result.
name / email