xuning / sherpaonnx

转到一个项目
Toggle navigation Toggle navigation pinning
Fangjun Kuang
Committed by GitHub
cd1fedaa 1 parent 69b6b47d

Add Dart API for audio tagging (#1181)

内嵌 并排对比
正在显示 30 个修改的文件 包含 504 行增加 和 18 行删除
@@ -4,6 +4,14 @@ set -ex @@ -4,6 +4,14 @@ set -ex
4 4
5 cd dart-api-examples 5 cd dart-api-examples
6 6
  7 +pushd audio-tagging
  8 +echo '----------zipformer----------'
  9 +./run-zipformer.sh
  10 +
  11 +echo '----------ced----------'
  12 +./run-ced.sh
  13 +popd
  14 +
7 pushd vad-with-non-streaming-asr 15 pushd vad-with-non-streaming-asr
8 echo '----------TeleSpeech CTC----------' 16 echo '----------TeleSpeech CTC----------'
9 ./run-telespeech-ctc.sh 17 ./run-telespeech-ctc.sh
@@ -110,6 +110,7 @@ jobs: @@ -110,6 +110,7 @@ jobs:
110 cp scripts/dart/tts-pubspec.yaml dart-api-examples/tts/pubspec.yaml 110 cp scripts/dart/tts-pubspec.yaml dart-api-examples/tts/pubspec.yaml
111 cp scripts/dart/kws-pubspec.yaml dart-api-examples/keyword-spotter/pubspec.yaml 111 cp scripts/dart/kws-pubspec.yaml dart-api-examples/keyword-spotter/pubspec.yaml
112 cp scripts/dart/vad-non-streaming-asr-pubspec.yaml dart-api-examples/vad-with-non-streaming-asr/pubspec.yaml 112 cp scripts/dart/vad-non-streaming-asr-pubspec.yaml dart-api-examples/vad-with-non-streaming-asr/pubspec.yaml
  113 + cp scripts/dart/audio-tagging-pubspec.yaml dart-api-examples/audio-tagging/pubspec.yaml
113 114
114 cp scripts/dart/sherpa-onnx-pubspec.yaml flutter/sherpa_onnx/pubspec.yaml 115 cp scripts/dart/sherpa-onnx-pubspec.yaml flutter/sherpa_onnx/pubspec.yaml
115 116
  1 +## 1.10.20
  2 +
  3 +* Add Dart API for audio tagging
  4 +
1 ## 1.10.19 5 ## 1.10.19
2 6
3 * Prefix all C API functions with SherpaOnnx 7 * Prefix all C API functions with SherpaOnnx
@@ -4,9 +4,13 @@ @@ -4,9 +4,13 @@
4 |------------------|------------------|----------------------|------------------------| 4 |------------------|------------------|----------------------|------------------------|
5 | ✔️ | ✔️ | ✔️ | ✔️ | 5 | ✔️ | ✔️ | ✔️ | ✔️ |
6 6
7 -| Spoken Language identification | Audio tagging | Voice activity detection | Keyword spotting |  
8 -|--------------------------------|---------------|--------------------------|------------------|  
9 -| ✔️ | ✔️ | ✔️ | ✔️ | 7 +| Spoken Language identification | Audio tagging | Voice activity detection |
  8 +|--------------------------------|---------------|--------------------------|
  9 +| ✔️ | ✔️ | ✔️ |
  10 +
  11 +| Keyword spotting | Add punctuation |
  12 +|------------------|-----------------|
  13 +| ✔️ | ✔️ |
10 14
11 ### Supported platforms 15 ### Supported platforms
12 16
@@ -5,7 +5,7 @@ This directory contains examples for Dart API. @@ -5,7 +5,7 @@ This directory contains examples for Dart API.
5 You can find the package at 5 You can find the package at
6 https://pub.dev/packages/sherpa_onnx 6 https://pub.dev/packages/sherpa_onnx
7 7
8 -## Descirption 8 +## Description
9 9
10 | Directory | Description | 10 | Directory | Description |
11 |-----------|-------------| 11 |-----------|-------------|
@@ -15,6 +15,7 @@ https://pub.dev/packages/sherpa_onnx @@ -15,6 +15,7 @@ https://pub.dev/packages/sherpa_onnx
15 | [./tts](./tts)| Example for text to speech| 15 | [./tts](./tts)| Example for text to speech|
16 | [./vad](./vad)| Example for voice activity detection| 16 | [./vad](./vad)| Example for voice activity detection|
17 | [./vad-with-non-streaming-asr](./vad-with-non-streaming-asr)| Example for voice activity detection with non-streaming speech recognition. You can use it to generate subtitles.| 17 | [./vad-with-non-streaming-asr](./vad-with-non-streaming-asr)| Example for voice activity detection with non-streaming speech recognition. You can use it to generate subtitles.|
  18 +| [./audio-tagging](./audio-tagging)| Example for audio tagging.|
18 19
19 ## How to create an example in this folder 20 ## How to create an example in this folder
20 21
  1 +# https://dart.dev/guides/libraries/private-files
  2 +# Created by `dart pub`
  3 +.dart_tool/
  1 +# Introduction
  2 +
  3 +This example shows how to use the Dart API from sherpa-onnx for audio tagging.
  4 +
  5 +| File | Description|
  6 +|------|------------|
  7 +|[./bin/zipformer.dart](./bin/zipformer.dart)| Use a Zipformer model for audio tagging. See [./run-zipformer.sh](./run-zipformer.sh)|
  8 +|[./bin/ced.dart](./bin/ced.dart)| Use a [CED](https://github.com/RicherMans/CED) model for audio tagging. See [./run-ced.sh](./run-ced.sh)|
  1 +# This file configures the static analysis results for your project (errors,
  2 +# warnings, and lints).
  3 +#
  4 +# This enables the 'recommended' set of lints from `package:lints`.
  5 +# This set helps identify many issues that may lead to problems when running
  6 +# or consuming Dart code, and enforces writing Dart using a single, idiomatic
  7 +# style and format.
  8 +#
  9 +# If you want a smaller set of lints you can change this to specify
  10 +# 'package:lints/core.yaml'. These are just the most critical lints
  11 +# (the recommended set includes the core lints).
  12 +# The core lints are also what is used by pub.dev for scoring packages.
  13 +
  14 +include: package:lints/recommended.yaml
  15 +
  16 +# Uncomment the following section to specify additional rules.
  17 +
  18 +# linter:
  19 +# rules:
  20 +# - camel_case_types
  21 +
  22 +# analyzer:
  23 +# exclude:
  24 +# - path/to/excluded/files/**
  25 +
  26 +# For more information about the core and recommended set of lints, see
  27 +# https://dart.dev/go/core-lints
  28 +
  29 +# For additional information about configuring this file, see
  30 +# https://dart.dev/guides/language/analysis-options
  1 +// Copyright (c) 2024 Xiaomi Corporation
  2 +import 'dart:io';
  3 +
  4 +import 'package:args/args.dart';
  5 +import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
  6 +import './init.dart';
  7 +
  8 +void main(List<String> arguments) async {
  9 + await initSherpaOnnx();
  10 +
  11 + final parser = ArgParser()
  12 + ..addOption('model', help: 'Path to the zipformer model')
  13 + ..addOption('labels', help: 'Path to class_labels_indices.csv')
  14 + ..addOption('top-k', help: 'topK events to be returned', defaultsTo: '5')
  15 + ..addOption('wav', help: 'Path to test.wav to be tagged');
  16 +
  17 + final res = parser.parse(arguments);
  18 + if (res['model'] == null || res['labels'] == null || res['wav'] == null) {
  19 + print(parser.usage);
  20 + exit(1);
  21 + }
  22 +
  23 + final model = res['model'] as String;
  24 + final labels = res['labels'] as String;
  25 + final topK = int.tryParse(res['top-k'] as String) ?? 5;
  26 + final wav = res['wav'] as String;
  27 +
  28 + final modelConfig = sherpa_onnx.AudioTaggingModelConfig(
  29 + ced: model,
  30 + numThreads: 1,
  31 + debug: true,
  32 + provider: 'cpu',
  33 + );
  34 +
  35 + final config = sherpa_onnx.AudioTaggingConfig(
  36 + model: modelConfig,
  37 + labels: labels,
  38 + );
  39 +
  40 + final at = sherpa_onnx.AudioTagging(config: config);
  41 +
  42 + final waveData = sherpa_onnx.readWave(wav);
  43 +
  44 + final stream = at.createStream();
  45 + stream.acceptWaveform(
  46 + samples: waveData.samples, sampleRate: waveData.sampleRate);
  47 +
  48 + final events = at.compute(stream: stream, topK: topK);
  49 +
  50 + print(events);
  51 +
  52 + stream.free();
  53 + at.free();
  54 +}
  1 +// Copyright (c) 2024 Xiaomi Corporation
  2 +import 'dart:io';
  3 +
  4 +import 'package:args/args.dart';
  5 +import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
  6 +import './init.dart';
  7 +
  8 +void main(List<String> arguments) async {
  9 + await initSherpaOnnx();
  10 +
  11 + final parser = ArgParser()
  12 + ..addOption('model', help: 'Path to the zipformer model')
  13 + ..addOption('labels', help: 'Path to class_labels_indices.csv')
  14 + ..addOption('top-k', help: 'topK events to be returned', defaultsTo: '5')
  15 + ..addOption('wav', help: 'Path to test.wav to be tagged');
  16 +
  17 + final res = parser.parse(arguments);
  18 + if (res['model'] == null || res['labels'] == null || res['wav'] == null) {
  19 + print(parser.usage);
  20 + exit(1);
  21 + }
  22 +
  23 + final model = res['model'] as String;
  24 + final labels = res['labels'] as String;
  25 + final topK = int.tryParse(res['top-k'] as String) ?? 5;
  26 + final wav = res['wav'] as String;
  27 +
  28 + final zipformerModelConfig =
  29 + sherpa_onnx.OfflineZipformerAudioTaggingModelConfig(
  30 + model: model,
  31 + );
  32 +
  33 + final modelConfig = sherpa_onnx.AudioTaggingModelConfig(
  34 + zipformer: zipformerModelConfig,
  35 + numThreads: 1,
  36 + debug: true,
  37 + provider: 'cpu',
  38 + );
  39 +
  40 + final config = sherpa_onnx.AudioTaggingConfig(
  41 + model: modelConfig,
  42 + labels: labels,
  43 + );
  44 +
  45 + final at = sherpa_onnx.AudioTagging(config: config);
  46 +
  47 + final waveData = sherpa_onnx.readWave(wav);
  48 +
  49 + final stream = at.createStream();
  50 + stream.acceptWaveform(
  51 + samples: waveData.samples, sampleRate: waveData.sampleRate);
  52 +
  53 + final events = at.compute(stream: stream, topK: topK);
  54 +
  55 + print(events);
  56 +
  57 + stream.free();
  58 + at.free();
  59 +}
  1 +name: audio_tagging
  2 +
  3 +description: >
  4 + This example demonstrates how to use the Dart API for audio tagging.
  5 +
  6 +version: 1.0.0
  7 +
  8 +environment:
  9 + sdk: ^3.4.0
  10 +
  11 +dependencies:
  12 + sherpa_onnx: ^1.10.19
  13 + path: ^1.9.0
  14 + args: ^2.5.0
  15 +
  16 +dev_dependencies:
  17 + lints: ^3.0.0
  1 +#!/usr/bin/env bash
  2 +
  3 +set -ex
  4 +
  5 +dart pub get
  6 +
  7 +if [[ ! -f ./sherpa-onnx-ced-mini-audio-tagging-2024-04-19/model.onnx ]]; then
  8 + curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/audio-tagging-models/sherpa-onnx-ced-mini-audio-tagging-2024-04-19.tar.bz2
  9 + tar xvf sherpa-onnx-ced-mini-audio-tagging-2024-04-19.tar.bz2
  10 + rm sherpa-onnx-ced-mini-audio-tagging-2024-04-19.tar.bz2
  11 +fi
  12 +
  13 +for w in 1 2 3 4 5 6; do
  14 + dart run \
  15 + ./bin/ced.dart \
  16 + --model ./sherpa-onnx-ced-mini-audio-tagging-2024-04-19/model.int8.onnx \
  17 + --labels ./sherpa-onnx-ced-mini-audio-tagging-2024-04-19/class_labels_indices.csv \
  18 + --wav ./sherpa-onnx-ced-mini-audio-tagging-2024-04-19/test_wavs/$w.wav
  19 +done
  1 +#!/usr/bin/env bash
  2 +
  3 +set -ex
  4 +
  5 +dart pub get
  6 +
  7 +if [[ ! -f ./sherpa-onnx-zipformer-audio-tagging-2024-04-09/model.onnx ]]; then
  8 + curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/audio-tagging-models/sherpa-onnx-zipformer-audio-tagging-2024-04-09.tar.bz2
  9 + tar xvf sherpa-onnx-zipformer-audio-tagging-2024-04-09.tar.bz2
  10 + rm sherpa-onnx-zipformer-audio-tagging-2024-04-09.tar.bz2
  11 +fi
  12 +
  13 +for w in 1 2 3 4 5 6; do
  14 + dart run \
  15 + ./bin/zipformer.dart \
  16 + --model ./sherpa-onnx-zipformer-audio-tagging-2024-04-09/model.int8.onnx \
  17 + --labels ./sherpa-onnx-zipformer-audio-tagging-2024-04-09/class_labels_indices.csv \
  18 + --wav ./sherpa-onnx-zipformer-audio-tagging-2024-04-09/test_wavs/$w.wav
  19 +done
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
@@ -65,5 +64,5 @@ void main(List<String> arguments) async { @@ -65,5 +64,5 @@ void main(List<String> arguments) async {
65 samples: audio.samples, 64 samples: audio.samples,
66 sampleRate: audio.sampleRate, 65 sampleRate: audio.sampleRate,
67 ); 66 );
68 - print('Saved to ${outputWav}'); 67 + print('Saved to $outputWav');
69 } 68 }
@@ -80,5 +80,5 @@ void main(List<String> arguments) async { @@ -80,5 +80,5 @@ void main(List<String> arguments) async {
80 samples: audio.samples, 80 samples: audio.samples,
81 sampleRate: audio.sampleRate, 81 sampleRate: audio.sampleRate,
82 ); 82 );
83 - print('Saved to ${outputWav}'); 83 + print('Saved to $outputWav');
84 } 84 }
1 // Copyright (c) 2024 Xiaomi Corporation 1 // Copyright (c) 2024 Xiaomi Corporation
2 import 'dart:io'; 2 import 'dart:io';
3 -import 'dart:typed_data';  
4 3
5 import 'package:args/args.dart'; 4 import 'package:args/args.dart';
6 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx; 5 import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
@@ -82,5 +81,5 @@ void main(List<String> arguments) async { @@ -82,5 +81,5 @@ void main(List<String> arguments) async {
82 samples: audio.samples, 81 samples: audio.samples,
83 sampleRate: audio.sampleRate, 82 sampleRate: audio.sampleRate,
84 ); 83 );
85 - print('Saved to ${outputWav}'); 84 + print('Saved to $outputWav');
86 } 85 }
@@ -77,5 +77,5 @@ void main(List<String> arguments) async { @@ -77,5 +77,5 @@ void main(List<String> arguments) async {
77 sherpa_onnx.writeWave( 77 sherpa_onnx.writeWave(
78 filename: outputWav, samples: s, sampleRate: waveData.sampleRate); 78 filename: outputWav, samples: s, sampleRate: waveData.sampleRate);
79 79
80 - print('Saved to ${outputWav}'); 80 + print('Saved to $outputWav');
81 } 81 }
@@ -2,6 +2,7 @@ @@ -2,6 +2,7 @@
2 import 'dart:io'; 2 import 'dart:io';
3 import 'dart:ffi'; 3 import 'dart:ffi';
4 4
  5 +export 'src/audio_tagging.dart';
5 export 'src/feature_config.dart'; 6 export 'src/feature_config.dart';
6 export 'src/keyword_spotter.dart'; 7 export 'src/keyword_spotter.dart';
7 export 'src/offline_recognizer.dart'; 8 export 'src/offline_recognizer.dart';
  1 +// Copyright (c) 2024 Xiaomi Corporation
  2 +import 'dart:ffi';
  3 +import 'package:ffi/ffi.dart';
  4 +
  5 +import './offline_stream.dart';
  6 +import './sherpa_onnx_bindings.dart';
  7 +
  8 +class OfflineZipformerAudioTaggingModelConfig {
  9 + const OfflineZipformerAudioTaggingModelConfig({this.model = ''});
  10 +
  11 + @override
  12 + String toString() {
  13 + return 'OfflineZipformerAudioTaggingModelConfig(model: $model)';
  14 + }
  15 +
  16 + final String model;
  17 +}
  18 +
  19 +class AudioTaggingModelConfig {
  20 + AudioTaggingModelConfig(
  21 + {this.zipformer = const OfflineZipformerAudioTaggingModelConfig(),
  22 + this.ced = '',
  23 + this.numThreads = 1,
  24 + this.provider = 'cpu',
  25 + this.debug = true});
  26 +
  27 + @override
  28 + String toString() {
  29 + return 'AudioTaggingModelConfig(zipformer: $zipformer, ced: $ced, numThreads: $numThreads, provider: $provider, debug: $debug)';
  30 + }
  31 +
  32 + final OfflineZipformerAudioTaggingModelConfig zipformer;
  33 + final String ced;
  34 + final int numThreads;
  35 + final String provider;
  36 + final bool debug;
  37 +}
  38 +
  39 +class AudioTaggingConfig {
  40 + AudioTaggingConfig({required this.model, this.labels = ''});
  41 +
  42 + @override
  43 + String toString() {
  44 + return 'AudioTaggingConfig(model: $model, labels: $labels)';
  45 + }
  46 +
  47 + final AudioTaggingModelConfig model;
  48 + final String labels;
  49 +}
  50 +
  51 +class AudioEvent {
  52 + AudioEvent({required this.name, required this.index, required this.prob});
  53 +
  54 + @override
  55 + String toString() {
  56 + return 'AudioEvent(name: $name, index: $index, prob: $prob)';
  57 + }
  58 +
  59 + final String name;
  60 + final int index;
  61 + final double prob;
  62 +}
  63 +
  64 +class AudioTagging {
  65 + AudioTagging._({required this.ptr, required this.config});
  66 +
  67 + // The user has to invoke AudioTagging.free() to avoid memory leak.
  68 + factory AudioTagging({required AudioTaggingConfig config}) {
  69 + final c = calloc<SherpaOnnxAudioTaggingConfig>();
  70 +
  71 + final zipformerPtr = config.model.zipformer.model.toNativeUtf8();
  72 + c.ref.model.zipformer.model = zipformerPtr;
  73 +
  74 + final cedPtr = config.model.ced.toNativeUtf8();
  75 + c.ref.model.ced = cedPtr;
  76 +
  77 + c.ref.model.numThreads = config.model.numThreads;
  78 +
  79 + final providerPtr = config.model.provider.toNativeUtf8();
  80 + c.ref.model.provider = providerPtr;
  81 +
  82 + c.ref.model.debug = config.model.debug ? 1 : 0;
  83 +
  84 + final labelsPtr = config.labels.toNativeUtf8();
  85 + c.ref.labels = labelsPtr;
  86 +
  87 + final ptr =
  88 + SherpaOnnxBindings.sherpaOnnxCreateAudioTagging?.call(c) ?? nullptr;
  89 +
  90 + calloc.free(labelsPtr);
  91 + calloc.free(providerPtr);
  92 + calloc.free(cedPtr);
  93 + calloc.free(zipformerPtr);
  94 + calloc.free(c);
  95 +
  96 + return AudioTagging._(ptr: ptr, config: config);
  97 + }
  98 +
  99 + void free() {
  100 + SherpaOnnxBindings.sherpaOnnxDestroyAudioTagging?.call(ptr);
  101 + ptr = nullptr;
  102 + }
  103 +
  104 + /// The user has to invoke stream.free() on the returned instance
  105 + /// to avoid memory leak
  106 + OfflineStream createStream() {
  107 + final p = SherpaOnnxBindings.sherpaOnnxAudioTaggingCreateOfflineStream
  108 + ?.call(ptr) ??
  109 + nullptr;
  110 + return OfflineStream(ptr: p);
  111 + }
  112 +
  113 + List<AudioEvent> compute({required OfflineStream stream, required int topK}) {
  114 + final pp = SherpaOnnxBindings.sherpaOnnxAudioTaggingCompute
  115 + ?.call(ptr, stream.ptr, topK) ??
  116 + nullptr;
  117 +
  118 + final ans = <AudioEvent>[];
  119 +
  120 + if (pp == nullptr) {
  121 + return ans;
  122 + }
  123 +
  124 + var i = 0;
  125 + while (pp[i] != nullptr) {
  126 + final p = pp[i];
  127 +
  128 + final name = p.ref.name.toDartString();
  129 + final index = p.ref.index;
  130 + final prob = p.ref.prob;
  131 + final e = AudioEvent(name: name, index: index, prob: prob);
  132 + ans.add(e);
  133 +
  134 + i += 1;
  135 + }
  136 +
  137 + SherpaOnnxBindings.sherpaOnnxAudioTaggingFreeResults?.call(pp);
  138 +
  139 + return ans;
  140 + }
  141 +
  142 + Pointer<SherpaOnnxAudioTagging> ptr;
  143 + final AudioTaggingConfig config;
  144 +}
@@ -2,6 +2,41 @@ @@ -2,6 +2,41 @@
2 import 'dart:ffi'; 2 import 'dart:ffi';
3 import 'package:ffi/ffi.dart'; 3 import 'package:ffi/ffi.dart';
4 4
  5 +final class SherpaOnnxOfflineZipformerAudioTaggingModelConfig extends Struct {
  6 + external Pointer<Utf8> model;
  7 +}
  8 +
  9 +final class SherpaOnnxAudioTaggingModelConfig extends Struct {
  10 + external SherpaOnnxOfflineZipformerAudioTaggingModelConfig zipformer;
  11 + external Pointer<Utf8> ced;
  12 +
  13 + @Int32()
  14 + external int numThreads;
  15 +
  16 + @Int32()
  17 + external int debug;
  18 +
  19 + external Pointer<Utf8> provider;
  20 +}
  21 +
  22 +final class SherpaOnnxAudioTaggingConfig extends Struct {
  23 + external SherpaOnnxAudioTaggingModelConfig model;
  24 + external Pointer<Utf8> labels;
  25 +
  26 + @Int32()
  27 + external int topK;
  28 +}
  29 +
  30 +final class SherpaOnnxAudioEvent extends Struct {
  31 + external Pointer<Utf8> name;
  32 +
  33 + @Int32()
  34 + external int index;
  35 +
  36 + @Float()
  37 + external double prob;
  38 +}
  39 +
5 final class SherpaOnnxOfflineTtsVitsModelConfig extends Struct { 40 final class SherpaOnnxOfflineTtsVitsModelConfig extends Struct {
6 external Pointer<Utf8> model; 41 external Pointer<Utf8> model;
7 external Pointer<Utf8> lexicon; 42 external Pointer<Utf8> lexicon;
@@ -303,6 +338,8 @@ final class SherpaOnnxKeywordSpotterConfig extends Struct { @@ -303,6 +338,8 @@ final class SherpaOnnxKeywordSpotterConfig extends Struct {
303 external Pointer<Utf8> keywordsFile; 338 external Pointer<Utf8> keywordsFile;
304 } 339 }
305 340
  341 +final class SherpaOnnxAudioTagging extends Opaque {}
  342 +
306 final class SherpaOnnxKeywordSpotter extends Opaque {} 343 final class SherpaOnnxKeywordSpotter extends Opaque {}
307 344
308 final class SherpaOnnxOfflineTts extends Opaque {} 345 final class SherpaOnnxOfflineTts extends Opaque {}
@@ -323,6 +360,40 @@ final class SherpaOnnxSpeakerEmbeddingExtractor extends Opaque {} @@ -323,6 +360,40 @@ final class SherpaOnnxSpeakerEmbeddingExtractor extends Opaque {}
323 360
324 final class SherpaOnnxSpeakerEmbeddingManager extends Opaque {} 361 final class SherpaOnnxSpeakerEmbeddingManager extends Opaque {}
325 362
  363 +typedef SherpaOnnxCreateAudioTaggingNative = Pointer<SherpaOnnxAudioTagging>
  364 + Function(Pointer<SherpaOnnxAudioTaggingConfig>);
  365 +
  366 +typedef SherpaOnnxCreateAudioTagging = SherpaOnnxCreateAudioTaggingNative;
  367 +
  368 +typedef SherpaOnnxDestroyAudioTaggingNative = Void Function(
  369 + Pointer<SherpaOnnxAudioTagging>);
  370 +
  371 +typedef SherpaOnnxDestroyAudioTagging = void Function(
  372 + Pointer<SherpaOnnxAudioTagging>);
  373 +
  374 +typedef SherpaOnnxAudioTaggingCreateOfflineStreamNative
  375 + = Pointer<SherpaOnnxOfflineStream> Function(
  376 + Pointer<SherpaOnnxAudioTagging>);
  377 +
  378 +typedef SherpaOnnxAudioTaggingCreateOfflineStream
  379 + = SherpaOnnxAudioTaggingCreateOfflineStreamNative;
  380 +
  381 +typedef SherpaOnnxAudioTaggingComputeNative
  382 + = Pointer<Pointer<SherpaOnnxAudioEvent>> Function(
  383 + Pointer<SherpaOnnxAudioTagging>,
  384 + Pointer<SherpaOnnxOfflineStream>,
  385 + Int32);
  386 +
  387 +typedef SherpaOnnxAudioTaggingCompute
  388 + = Pointer<Pointer<SherpaOnnxAudioEvent>> Function(
  389 + Pointer<SherpaOnnxAudioTagging>, Pointer<SherpaOnnxOfflineStream>, int);
  390 +
  391 +typedef SherpaOnnxAudioTaggingFreeResultsNative = Void Function(
  392 + Pointer<Pointer<SherpaOnnxAudioEvent>>);
  393 +
  394 +typedef SherpaOnnxAudioTaggingFreeResults = void Function(
  395 + Pointer<Pointer<SherpaOnnxAudioEvent>>);
  396 +
326 typedef CreateKeywordSpotterNative = Pointer<SherpaOnnxKeywordSpotter> Function( 397 typedef CreateKeywordSpotterNative = Pointer<SherpaOnnxKeywordSpotter> Function(
327 Pointer<SherpaOnnxKeywordSpotterConfig>); 398 Pointer<SherpaOnnxKeywordSpotterConfig>);
328 399
@@ -804,6 +875,13 @@ typedef SherpaOnnxFreeWaveNative = Void Function(Pointer<SherpaOnnxWave>); @@ -804,6 +875,13 @@ typedef SherpaOnnxFreeWaveNative = Void Function(Pointer<SherpaOnnxWave>);
804 typedef SherpaOnnxFreeWave = void Function(Pointer<SherpaOnnxWave>); 875 typedef SherpaOnnxFreeWave = void Function(Pointer<SherpaOnnxWave>);
805 876
806 class SherpaOnnxBindings { 877 class SherpaOnnxBindings {
  878 + static SherpaOnnxCreateAudioTagging? sherpaOnnxCreateAudioTagging;
  879 + static SherpaOnnxDestroyAudioTagging? sherpaOnnxDestroyAudioTagging;
  880 + static SherpaOnnxAudioTaggingCreateOfflineStream?
  881 + sherpaOnnxAudioTaggingCreateOfflineStream;
  882 + static SherpaOnnxAudioTaggingCompute? sherpaOnnxAudioTaggingCompute;
  883 + static SherpaOnnxAudioTaggingFreeResults? sherpaOnnxAudioTaggingFreeResults;
  884 +
807 static CreateKeywordSpotter? createKeywordSpotter; 885 static CreateKeywordSpotter? createKeywordSpotter;
808 static DestroyKeywordSpotter? destroyKeywordSpotter; 886 static DestroyKeywordSpotter? destroyKeywordSpotter;
809 static CreateKeywordStream? createKeywordStream; 887 static CreateKeywordStream? createKeywordStream;
@@ -958,6 +1036,33 @@ class SherpaOnnxBindings { @@ -958,6 +1036,33 @@ class SherpaOnnxBindings {
958 static SherpaOnnxFreeWave? freeWave; 1036 static SherpaOnnxFreeWave? freeWave;
959 1037
960 static void init(DynamicLibrary dynamicLibrary) { 1038 static void init(DynamicLibrary dynamicLibrary) {
  1039 + sherpaOnnxCreateAudioTagging ??= dynamicLibrary
  1040 + .lookup<NativeFunction<SherpaOnnxCreateAudioTaggingNative>>(
  1041 + 'SherpaOnnxCreateAudioTagging')
  1042 + .asFunction();
  1043 +
  1044 + sherpaOnnxDestroyAudioTagging ??= dynamicLibrary
  1045 + .lookup<NativeFunction<SherpaOnnxDestroyAudioTaggingNative>>(
  1046 + 'SherpaOnnxDestroyAudioTagging')
  1047 + .asFunction();
  1048 +
  1049 + sherpaOnnxAudioTaggingCreateOfflineStream ??= dynamicLibrary
  1050 + .lookup<
  1051 + NativeFunction<
  1052 + SherpaOnnxAudioTaggingCreateOfflineStreamNative>>(
  1053 + 'SherpaOnnxAudioTaggingCreateOfflineStream')
  1054 + .asFunction();
  1055 +
  1056 + sherpaOnnxAudioTaggingCompute ??= dynamicLibrary
  1057 + .lookup<NativeFunction<SherpaOnnxAudioTaggingComputeNative>>(
  1058 + 'SherpaOnnxAudioTaggingCompute')
  1059 + .asFunction();
  1060 +
  1061 + sherpaOnnxAudioTaggingFreeResults ??= dynamicLibrary
  1062 + .lookup<NativeFunction<SherpaOnnxAudioTaggingFreeResultsNative>>(
  1063 + 'SherpaOnnxAudioTaggingFreeResults')
  1064 + .asFunction();
  1065 +
961 createKeywordSpotter ??= dynamicLibrary 1066 createKeywordSpotter ??= dynamicLibrary
962 .lookup<NativeFunction<CreateKeywordSpotterNative>>( 1067 .lookup<NativeFunction<CreateKeywordSpotterNative>>(
963 'SherpaOnnxCreateKeywordSpotter') 1068 'SherpaOnnxCreateKeywordSpotter')
  1 +name: audio_tagging
  2 +
  3 +description: >
  4 + This example demonstrates how to use the Dart API for audio tagging.
  5 +
  6 +version: 1.0.0
  7 +
  8 +environment:
  9 + sdk: ^3.4.0
  10 +
  11 +dependencies:
  12 + sherpa_onnx:
  13 + path: ../../flutter/sherpa_onnx
  14 + path: ^1.9.0
  15 + args: ^2.5.0
  16 +
  17 +dev_dependencies:
  18 + lints: ^3.0.0