Respiratory sound classification on ICBHI dataset official (60-40% split)
72.7ScoreResp-Agent
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Resp-AgentBackbone=LLM+Longformer, Pretraining Data=HF+SPR2026.02 | 72.7 | 79.29 | 66.1 | |
| Dong et al.Backbone=AST, Pretraining Data=IN+AS2026.02 | 67.55 | 85.99 | 49.11 | |
| MVSTBackbone=AST, Pretraining Data=IN+AS2026.02 | 66.55 | 81.99 | 51.1 | |
| Meta-EnsembleBackbone=CLAP, Pretraining Data=LA, Split Strategy=Fixed Split, Level=S-level, Architecture=2-Hidden2026.04 | 66.49 | 89.6 | 43.54 | |
| Mean-EnsembleBackbone=CLAP, Pretraining Data=LA, Split Strategy=Fixed Split, Level=S-level2026.04 | 66.34 | 89.87 | 42.82 | |
| Toikkanen et al. (BTS++)Backbone=CLAP, Pretraining Data=LA, Venue=INTERSPEECH‘252026.04 | 65.69 | 89.49 | 41.89 | |
| Meta-EnsembleBackbone=CLAP, Pretraining Data=LA, Split Strategy=5-Fold, Level=S-level, Architecture=2-Hidden2026.04 | 65.63 | 86.16 | 45.1 | |
| Toikkanen et al. (BTS-d)Backbone=CLAP, Pretraining Data=LA, Venue=INTERSPEECH‘252026.04 | 64.39 | 82.89 | 45.9 | |
| Mean-EnsembleBackbone=CLAP, Pretraining Data=LA, Split Strategy=5-Fold, Level=P-level2026.04 | 64.33 | 87.21 | 41.46 | |
| Wang et al. (2024)Backbone=HTS-AT, Pretraining Data=IN+AS2026.02 | 64.19 | 79.61 | 48.77 | |
| BTSBackbone=CLAP, Pretraining Data=LA2024.06 | 63.54 | 81.4 | 45.67 | |
| BTSBackbone=CLAP, Pretraining Data=LA2026.02 | 63.54 | 81.4 | 45.67 | |
| Kim et al. (BTS)Backbone=CLAP, Pretraining Data=LA, Venue=INTERSPEECH‘242026.04 | 63.54 | 81.4 | 45.67 | |
| Kim et al. (BTS)Backbone=CLAP, Pretraining Data=LA2026.06 | 63.54 | 81.4 | 45.67 | |
| QLung on Audio-CLAPBackbone=CLAP, Pretraining Data=LA2026.06 | 63.39 | 81.98 | 44.81 | |
| Daisuke et al. (M2D-X/0.7)Backbone=M2D ViT, Pretraining Data=AS, Venue=TASLP‘242026.04 | 63.29 | 81.51 | 45.08 | |
| Audio-CLAPBackbone=CLAP, Pretraining Data=LA2024.06 | 62.56 | 80.85 | 44.67 | |
| Kim et al. (Audio-CLAP)Backbone=CLAP, Pretraining Data=LA, Venue=INTERSPEECH‘242026.04 | 62.56 | 80.85 | 44.67 | |
| Kim et al. (Audio-CLAP)Backbone=CLAP, Pretraining Data=LA2026.06 | 62.56 | 80.85 | 44.67 | |
| Xiao et al. (LungAdapter)Backbone=AST, Pretraining Data=IN + AS2026.06 | 62.4 | 80.43 | 44.37 | |
| Bae et al. (Patch-Mix CL)architecture=AST, pretrain=IN + AS2023.11 | 62.37 | 81.66 | 43.07 | |
| Bae et al. (Patch-Mix CL)Backbone=AST, Pretraining Data=IN + AS2024.06 | 62.37 | 81.66 | 43.07 | |
| Bae et al.Backbone=AST, Pretraining Data=IN+AS2026.02 | 62.37 | 81.66 | 43.07 | |
| Bae et al. (Patch-Mix CL)Backbone=AST, Pretraining Data=IN + AS, Venue=INTERSPEECH‘232026.04 | 62.37 | 81.66 | 43.07 | |
| Bae et al. (Patch-Mix CL)Backbone=AST, Pretraining Data=IN + AS2026.06 | 62.37 | 81.66 | 43.07 | |
| QLung on ASTBackbone=AST, Pretraining Data=IN + AS2026.06 | 62.01 | 81.9 | 42.12 | |
| Kim et al. (AFT on Mixed-500)Backbone=AST, Pretraining Data=IN + AS2024.06 | 61.79 | 80.72 | 42.86 | |
| Kim et al. (2023)Backbone=AST, Pretraining Data=IN+AS2026.02 | 61.79 | 80.72 | 42.86 | |
| Kim et al. (SG-SCL)Backbone=AST, Pretraining Data=IN + AS2024.06 | 61.71 | 79.87 | 43.55 | |
| Kim et al. (2024a)Backbone=AST, Pretraining Data=IN+AS2026.02 | 61.71 | 79.87 | 43.55 | |
| Kim et al. (SG-SCL)Backbone=AST, Pretraining Data=IN + AS, Venue=ICASSP‘242026.04 | 61.71 | 79.87 | 43.55 | |
| Kim et al. (SG-SCL)Backbone=AST, Pretraining Data=IN + AS2026.06 | 61.71 | 79.87 | 43.55 | |
| Kim et al. (RepAugment)Backbone=AST, Pretraining Data=IN + AS2024.06 | 61.51 | 82.47 | 40.55 | |
| Kim et al. (2024b)Backbone=AST, Pretraining Data=IN+AS2026.02 | 61.51 | 82.47 | 40.55 | |
| Bae et al. (Fine-tuning)architecture=AST, pretrain=IN + AS2023.11 | 59.55 | 77.14 | 41.97 | |
| Bae et al. (Fine-tuning)Backbone=AST, Pretraining Data=IN + AS2024.06 | 59.55 | 77.14 | 41.97 | |
| Bae et al. (Fine-tuning)Backbone=AST, Pretraining Data=IN + AS2026.06 | 59.55 | 77.14 | 41.97 | |
| Ge et al. (Lungmix)Backbone=AST, Pretraining Data=IN + AS2026.06 | 58.53 | — | — | |
| Nguyen et al. (CoTuning)architecture=ResNet50, pretrain=IN2023.11 | 58.29 | 79.34 | 37.24 | |
| Nguyen et al. (CoTuning)Backbone=ResNet50, Pretraining Data=IN2024.06 | 58.29 | 79.34 | 37.24 | |
| Nguyen et al.Backbone=ResNet50, Pretraining Data=IN2026.02 | 58.29 | 79.34 | 37.24 | |
| Nguyen et al. (StochNorm)Backbone=ResNet50, Pretraining Data=IN2024.06 | 57.63 | 78.86 | 36.4 | |
| Moummad et al. (SCL)architecture=CNN6, pretrain=AS2023.11 | 57.55 | 75.95 | 39.15 | |
| Moummad et al. (SCL)Backbone=CNN6, Pretraining Data=AS2024.06 | 57.55 | 75.95 | 39.15 | |
| Moummad et al.Backbone=CNN6, Pretraining Data=AS2026.02 | 57.55 | 75.95 | 39.15 | |
| Late-FusionBackbone=Inc-03 + VGG14, Pretraining Data=IN2024.06 | 57.3 | 85.6 | 30 | |
| Late-FusionBackbone=Inc-03 + VGG14, Pretraining Data=IN2026.02 | 57.3 | 85.6 | 30 | |
| RespireNet (CBA+BRC+FT)architecture=ResNet34, pretrain=IN2023.11 | 56.2 | 72.3 | 40.1 | |
| RespireNet (CBA+BRC+FT)Backbone=ResNet34, Pretraining Data=IN2024.06 | 56.2 | 72.3 | 40.1 | |
| RespireNetBackbone=ResNet34, Pretraining Data=IN2026.02 | 56.2 | 72.3 | 40.1 | |
| Wang et al. (Splice)architecture=ResNeSt, pretrain=IN2023.11 | 55.3 | 70.4 | 40.2 | |
| Wang et al. (Splice)Backbone=ResNeSt, Pretraining Data=IN2024.06 | 55.3 | 70.4 | 40.2 | |
| Wang et al.Backbone=ResNeSt, Pretraining Data=IN2026.02 | 55.3 | 70.4 | 40.2 | |
| Wang et al. (Splice)Backbone=ResNeSt, Pretraining Data=IN2026.06 | 55.3 | 70.4 | 40.2 | |
| Moummad et al.Backbone=CNN6, Pretraining Data=AS2024.06 | 55.24 | 70.09 | 40.39 | |
| Chang et al.Backbone=CNN8-dilated2024.06 | 52.89 | 69.92 | 35.85 | |
| Chang et al.Backbone=CNN8-dilated, Pretraining Data=-2026.02 | 52.89 | 69.92 | 35.85 | |
| Chang et al.Backbone=CNN8-dilated, Pretraining Data=-, Venue=INTERSPEECH‘222026.04 | 52.89 | 69.92 | 35.85 | |
| LungRN+NLBackbone=ResNet-NL2024.06 | 52.26 | 63.2 | 41.32 | |
| LungRN+NLBackbone=ResNet-NL, Pretraining Data=-2026.02 | 52.26 | 63.2 | 41.32 | |
| LungRN+NLBackbone=ResNet-NL, Pretraining Data=-2026.06 | 52.26 | 63.2 | 41.32 | |
| Ren et al.Backbone=CNN8-Pt2024.06 | 50.37 | 72.96 | 27.78 | |
| Ren et al.Backbone=CNN8-Pt, Pretraining Data=-2026.02 | 50.37 | 72.96 | 27.78 | |
| Ren et al.Backbone=CNN8-Pt, Pretraining Data=-2026.06 | 50.37 | 72.96 | 27.78 | |
| SE+SABackbone=ResNet182024.06 | 49.55 | 81.25 | 17.84 | |
| SE+SABackbone=ResNet18, Pretraining Data=-2026.02 | 49.55 | 81.25 | 17.84 | |
| SE+SABackbone=ResNet18, Pretraining Data=-2026.06 | 49.55 | 81.25 | 17.84 |