Views
No views yet
[!IMPORTANT] Official model for our INTERSPEECH 2026 paper "A Data-Centric Framework for Addressing Phonetic and Prosodic Challenges in Russian Speech Generative Models" (arXiv:2507.13563). Part of the Balalaika Russian speech data-processing pipeline — code: https://github.com/lab260ru/balalaika. If you use this resource, please cite it.
microsoft/wavlm-base-plus
the best threshold value 0.2442git clone https://huggingface.co/MTUCI/MusicDetection
cd MusicDetection
pip install -r requirements.txt1from model import WavLMForMusicDetection
2from safetensors import safe_open
3
4model = WavLMForMusicDetection(batch_size=32, device='cuda')
5with safe_open('music_detection.safetensors', framework="pt") as f:
6 model.load_state_dict({k: f.get_tensor(k) for k in f.keys()})
7
8probs = model.predict_proba(['audio1.mp3', 'audio2.wav']) # → tensor([0.88, 0.11])
9
10## Contact
11
12- Email: kborodin.research@gmail.com
13- Telegram: [@korallll_ai](https://t.me/korallll_ai)
14
15## Citation
16
17If you use this resource, please cite our INTERSPEECH 2026 paper:
18
19```bibtex
20@inproceedings{borodin2026balalaika,
21 title = {A Data-Centric Framework for Addressing Phonetic and Prosodic Challenges in Russian Speech Generative Models},
22 author = {Borodin, Kirill and Vasiliev, Nikita and Kudryavtsev, Vasiliy and Maslov, Maxim and Gorodnichev, Mikhail and Rogov, Oleg and Mkrtchian, Grach},
23 booktitle = {Proc. INTERSPEECH 2026},
24 year = {2026},
25 note = {arXiv:2507.13563},
26 url = {https://arxiv.org/abs/2507.13563}
27}