2024:Music Audio Generation Results

From MIREX Wiki
Revision as of 11:36, 11 November 2024 by Junyan (talk | contribs) (Created page with "= Submissions = {| class="wikitable" style="text-align:center;" |- style="font-weight:bold; vertical-align:bottom;" ! style="text-align:left;" | Team ! style="text-align:left...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Submissions

Team Extended Abstract Methods Used
S1-CodecLM 7B decoder only from scratch + 2 stage semantic tokenizer
B1-MusicGen-Large PDF MusicGen
B2-MusicGen-Medium PDF MusicGen
B3-MusicGen-Small PDF MusicGen

Results

Team Frechet Distance ↓ Frechet Audio Distance ↓ Kullback Leibler Divergence ↓ Inception Score ↑ Relative Overall ↑ (Normalized Avg)
S1-CodecLM 13.77 2.67 1.71 1.52±0.04 0.716
B1-MusicGen-Large 19.05 2.5 2.11 1.57±0.03 0.672
B2-MusicGen-Medium 24.58 3.59 2.46 1.61±0.06 0.356
B3-MusicGen-Small 26.21 3.75 2.61 1.58±0.04 0.167
Ground-truth 0 0 0 1.65±0.06 -