Skip to content

Yamnet for speech classification using CPP and ONNX-runtime-2025高通边缘智能创新应用大赛入围决赛方案

Notifications You must be signed in to change notification settings

Choise-ieee/yamnet_onnx_cpp_audio_speech_classification

Repository files navigation

yamnet onnx cpp audio sppech classifcation

2025高通边缘智能创新应用大赛入围决赛方案-智能终端赛道[https://qc-ieiot-challenge.aidlux.com/2025/]

Finalist Proposal for the 2025 Qualcomm Edge Intelligence Innovation Application Competition - Smart Terminal Track

Coding info

the coding can adaptive the windows and Linux system. The running time for quallcomm QCM need onnx envirment, we can use the following command to build it:

  1. Download the onnx version 1.181, the windows select the X86 version and GPU version,the Qualcomm select the ARM version, and the Linux select the aarch Linux version
  2. In Qualcomm environment, put the relative lib file libonnxruntime.so and so on to /usr/lib
  3. choose the input wav file in filepath.
  4. select topk=5 or topk=10, to show the result of topk

Caution

The wav file generated by FFMPEG should satisfy:

  1. -bitexact
  2. MAX_SAMPLES = 48000
  3. SAMPLE_RATE = 16000;
  4. CHANNELS = 1;

Demo in windows

wechat_2025-10-06_091545_319

Demo in Qualcomm Linux

The yamnet in this project is used for analyzing the speech detect firstly, and then used to sync detect(https://github.com/Choise-ieee/video-and-audio-sync-detection-based-on-lips) image

Thanks the 2025 Qualcomm Edge Intelligence Innovation Application Competition, Aidlux,Quectel ,AidLux_Me(阿加犀小助手)and so on。

About

Yamnet for speech classification using CPP and ONNX-runtime-2025高通边缘智能创新应用大赛入围决赛方案

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages