Uploaded image for project: 'Picard'
  1. Picard
  2. PICARD-827

Improved Audio Fingerprinting for clusters

    XMLWordPrintable

    Details

      Description

      I posted this in the forum first, https://community.metabrainz.org/t/feature-request-for-better-fingerprint-scanning-and-album-matching-from-a-cluster/79528

      I think my request requires two parts, a new webservice and a (change/choice) to the way Picard scans a cluster.

      I don't know about other people but if you scan a cluster to use the audio fingerprinting information one album can get smeared across a number of different albums, sometimes nearly as many as there are tracks.

      I think it would work better if it did this, scan a cluster and get all the possible recording mbid's from Acoustid for each track, not just the first one. Do one batch submission to a new MusicBrainz web service that contains all the possible mbid's for all songs in the cluster then using a query like the one below MB would return the most likely release mbid by number of tracks contained.

      I have downloaded the MusicBrainz database and found this method is not only quick and accurate but also reduces the number of web-service calls dramatically.

      SQL query

      SELECT TOP 5 rl.gid,
      rl.NAME,
      Count(rc.gid) AS [COUNT]
      FROM recording rc
      JOIN track t
      ON rc.id = t.recording
      JOIN medium m
      ON m.id = t.medium
      JOIN release rl
      ON rl.id = m.release
      WHERE rc.gid IN ( GUID, GUID, GUID, GUID, GUID, GUID, etc . . . )
      GROUP BY rl.gid,
      rl.NAME
      ORDER BY 3 DESC

        Attachments

          Activity

            People

            • Assignee:
              Unassigned
              Reporter:
              matt_heys Matt
            • Votes:
              3 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

              • Created:
                Updated:

                Packages

                Version Package