Segment-level Video-to-Music Retrieval on MovieClips (test)
21Median RankMVPt
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MVPtArchitecture=Transformer, Audio Features=DeepSim, Visual Features=CLIP, Loss Function=InfoNCE2022.06 | 21 | 0.1508 | 0.2555 | 0.3625 | 0.3656 | |
| MVPt + X3D featuresArchitecture=Transformer, Audio Features=DeepSim, Visual Features=CLIP + X3D, Loss Function=InfoNCE2022.06 | 28 | 0.0858 | 0.1908 | 0.2852 | 0.2911 | |
| Baseline + DeepSim + CLIP + InfoNCEArchitecture=MLP, Audio Features=DeepSim, Visual Features=CLIP, Loss Function=InfoNCE2022.06 | 74 | 0.0253 | 0.0795 | 0.1499 | 0.1942 | |
| Baseline + DeepSim + CLIPArchitecture=MLP, Audio Features=DeepSim, Visual Features=CLIP, Loss Function=Triplet2022.06 | 189 | 0.021 | 0.058 | 0.0936 | 0.1135 | |
| Chance2022.06 | 1,000 | 0.0005 | 0.0025 | 0.005 | 0.005 |