Downloads · 30 days
5
7% of all-time downloads
Wedefense_ASV2025_WavLM_Base_Pruning
JYP2024/Wedefense_ASV2025_WavLM_Base_Pruning
Wedefense_ASV2025_WavLM_Base_Pruning is a audio classification model from JYP2024. Use it for the audio classification task on the model card, and read the license before you ship it in a product.
- Input Feature: Raw waveform (via SSL model) - Frame Configuration: 150 frames per segment, 20 ms frame shift - Training Strategy: Jointly optimizing for task performance and model sparsity in a single stage. A warm-…
All-time downloads
67
Public
Repo size
1.1 GB
Likes
2
Public
Repo makeup
Click a slice to open those files.
.bin684 MB · 63%
At a glance
- Task
- Audio Classification
- Access
- Public
- Created
- Aug 11, 2025
- Updated
- Aug 12, 2025
- SHA
- 6993ddda
Base models
Datasets
Model card
From the Hugging Face model README
🔎 Hybrid Pruning for Anti-Spoofing Results
- Input Feature: Raw waveform (via SSL model)
- Frame Configuration: 150 frames per segment, 20 ms frame shift
- Training Strategy: Jointly optimizing for task performance and model sparsity in a single stage. A warm-up schedule is used where the sparsity target linearly increases from 0 to the final value over the first 5 epochs.
- Evaluation Metrics: minDCF, EER (%)
- Evaluation Sets: Dev / Eval
- Back-end: Multi-Head Factorized Attentive Pooling (MHFA)
Results on ASVspoof 5
The following table compares the performance of our proposed Hybrid Pruning (HP) single system against other top-performing systems from the official ASVspoof 5 Challenge leaderboard.
| System | Dev minDCF | Dev EER (%) | Eval minDCF | Eval EER (%) |
|---|---|---|---|---|
| Rank 3 (ID:T27, Fusion) | - | - | 0.0937 | 3.42 |
| HP (ours, Single system) | 0.0395 | 1.547 | 0.1028 | 3.758 |
| Rank 4 (ID:T23, Fusion) | - | - | 0.1124 | 4.16 |
| Rank 9 (ID:T23, Best single system) | - | - | 0.1499 | 5.56 |
- Task
- Audio Classification
- Languages
- en
- Created
- Aug 11, 2025
- Updated
- Aug 12, 2025