A Implementation of SpecAugment with Tensorflow & Pytorch, introduced by Google Brain

🔦 A Pytorch implementation of GoogleBrain's SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

Tensor2tensor experiment with SpecAugment

A Pytorch implementation of the paper : SpecAugment++: A Hidden Space Data Augmentation Method for Acoustic Scene Classification

End-to-end speech recognition on AISHELL dataset.

fast SpecAugmentation code with numpy and scipy

Pytorch implementation of sparse_image_warp and an example of GoogleBrain's SpecAugment is given: A Simple Data Augmentation Method for Automatic Speech Recognition https://arxiv.org/abs/1904.08779

tf 2.0 implementation of Listen, attend and spell

A KALDI/C++ implementation of GoogleBrain's SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

A Pytorch (support batch and channel) implementation of GoogleBrain's SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

PyTorch Implementation of Time/Frequency Masks

EEG-Epilepsy-Prediction 是一个面向多通道脑电（EEG）数据的癫痫发作检测与分型框架，集成了频谱/时域特征提取、BiLSTM 帧级分类与精细后处理（平滑、阈值判决、确认窗、冷却合并），聚焦事件级检测精度与误报控制（FA/h 限制）。项目支持 特征缓存与标签自动构建，并提供 阈值网格搜索+写回配置，实现端到端训练与推理。创新点包括：引入 Mixup/SpecAugment 数据增强 提升模型泛化，采用 LOSOCV（受试者留一）策略 评估跨个体鲁棒性，结合轻量化实现与参数寻优工具，适合科研探索与临床原型开发。

Tensorflow Layer that implements the SpecAugment technique

A minimalistic Tensorflow 2.x Keras layer which applies SpecAugment to its input

对谷歌最新的论文SpecAugment使用matlab复现

Emotion recognition with IEMOCAP datasets. We compare the results with SpecAugmentation and CodecAugmentation. For audio codec implementation, we have selected opus.

HCMUS at MediaEval 2020: Emotion Classification Using Wavenet Features with SpecAugment and EfficientNet

Emotion recognition with IEMOCAP datasets. We compare the results with SpecAugmentation and CodecAugmentation. For audio codec implementation, we have selected opus.

sparse_image_warp module supporting dynamic shape tensor and time warp function for specaugment

SpecAugment data augmentation + Wav2Vec2 fine-tuning for dementia speech classification

[Naver AI hackathon: Speech Recognition] Rank 22/100: LAS + Multihead attention + Silence trimming + SpecAugment + Scheduled sampling + LR decay + N-gram LM + Beam search

XSpeech: A Novel Deep Learning Approach to Classifying Stutters

SagalNet is a robust end-to-end Machine Learning pipeline for converting spoken Afaan Oromoo digits into text. Features a custom DeeperCNN architecture, SpecAugment for noise resilience, MLflow experiment tracking, and an interactive Streamlit UI for live inference.

A ResNet‑Audio pipeline detects synthetic speech across five languages with calibrated confidence. Using SE‑ResNet attention, it balances pitch and spectral cues, applies sliding‑window inference over 5‑second chunks, and calibrates scores with temperature scaling. Robustness is ensured through training with telephony simulation and SpecAugment.

SpecAugment implementation.

No description available

No description available

No description available

No description available