Text-to-Audio Retrieval on Soundscape Datasets (HSN, NES, SNE, UHH, PER, SSW) 2024 (OOD)
40.69HSN Scoretext-observation hashing framework
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| text-observation hashing frameworkHashing Bits=128-bit, Representation Type=Continuous (LoRA)2026.01 | 40.69 | 29.94 | 30.48 | 28.3 | 9.96 | 49.15 | 31.42 | |
| text-observation hashing frameworkHashing Bits=256-bit, Representation Type=Binary (Hashing)2026.01 | 39.65 | 28.8 | 34.02 | 29.66 | 8.7 | 55.75 | 32.76 | |
| text-observation hashing frameworkHashing Bits=256-bit, Representation Type=Continuous (LoRA)2026.01 | 38.11 | 28.26 | 28.91 | 33 | 11.1 | 43.84 | 30.53 | |
| text-observation hashing frameworkHashing Bits=128-bit, Representation Type=Binary (Hashing)2026.01 | 35.2 | 28.28 | 34 | 29.77 | 6.89 | 55.3 | 31.57 | |
| BioLingualFeature=Original Embeddings2026.01 | 27.85 | 24.32 | 21.49 | 32 | 9.73 | 33.64 | 24.83 |