Music-to-Video Retrieval on MusicVid-YT8M track-level (test)
13Median RankMVPt
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MVPtaudio features=DeepSim, visual features=CLIP, temporal model=Transformer, loss=InfoNCE2022.06 | 13 | 6.36 | 25.73 | 42.65 | 42.27 | |
| Baseline + DeepSim + CLIP + Transformers + Temporal embeddingsaudio features=DeepSim, visual features=CLIP, temporal model=Transformer, positional encoding=true2022.06 | 17 | 6.22 | 24.43 | 40.68 | 39.5 | |
| Baseline + DeepSim + CLIP + Transformers w/o positionaudio features=DeepSim, visual features=CLIP, temporal model=Transformer, positional encoding=false2022.06 | 21 | 5.86 | 21.81 | 35.7 | 34.05 | |
| Baseline + DeepSim + CLIP visual featuresaudio features=DeepSim, visual features=CLIP2022.06 | 45 | 5.03 | 15.7 | 24.09 | 21.33 | |
| Baseline + DeepSim audio featuresaudio features=DeepSim2022.06 | 94 | 2.29 | 8.55 | 13.8 | 11.41 | |
| Baseline2022.06 | 98 | 2.57 | 8.61 | 13.81 | 9.86 | |
| Chance2022.06 | 1,000 | 0.05 | 0.25 | 0.5 | 0.5 |