talksik/mediapipe
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Small TS audio API improvement

Browse Source 
PiperOrigin-RevId: 490374083
This commit is contained in:
MediaPipe Team
2022-11-22 17:27:47 -08:00
committed by Copybara-Service
parent efa9e737f8
commit fac97554df
3 changed files with 105 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mediapipe/tasks/web/audio/audio_classifier/audio_classifier.ts
+2 -12
View File
@@ -35,11 +35,7 @@ export * from './audio_classifier_result';
const MEDIAPIPE_GRAPH =
'mediapipe.tasks.audio.audio_classifier.AudioClassifierGraph';
// Note: `input_audio` is hardcoded in 'gl_graph_runner_internal_audio' and
// cannot be changed
// TODO: Change this to `audio_in` to match the name in the CC
// implementation
const AUDIO_STREAM = 'input_audio';
const AUDIO_STREAM = 'audio_in';
const SAMPLE_RATE_STREAM = 'sample_rate';
const TIMESTAMPED_CLASSIFICATIONS_STREAM = 'timestamped_classifications';
@@ -154,14 +150,8 @@ export class AudioClassifier extends AudioTaskRunner<AudioClassifierResult[]> {
protected override process(
audioData: Float32Array, sampleRate: number,
timestampMs: number): AudioClassifierResult[] {
// Configures the number of samples in the WASM layer. We re-configure the
// number of samples and the sample rate for every frame, but ignore other
// side effects of this function (such as sending the input side packet and
// the input stream header).
this.configureAudio(
/* numChannels= */ 1, /* numSamples= */ audioData.length, sampleRate);
this.addDoubleToStream(sampleRate, SAMPLE_RATE_STREAM, timestampMs);
this.addAudioToStream(audioData, timestampMs);
this.addAudioToStreamWithShape(audioData, /* numChannels= */ 1, /* numSamples= */ audioData.length, AUDIO_STREAM, timestampMs);
this.classificationResults = [];
this.finishProcessing();
mediapipe/tasks/web/audio/audio_embedder/audio_embedder.ts
+2 -12
View File
@@ -35,11 +35,7 @@ export * from './audio_embedder_result';
// The OSS JS API does not support the builder pattern.
// tslint:disable:jspb-use-builder-pattern
// Note: `input_audio` is hardcoded in 'gl_graph_runner_internal_audio' cannot
// be changed
// TODO: Change this to `audio_in` to match the name in the CC
// implementation
const AUDIO_STREAM = 'input_audio';
const AUDIO_STREAM = 'audio_in';
const SAMPLE_RATE_STREAM = 'sample_rate';
const EMBEDDINGS_STREAM = 'embeddings_out';
const TIMESTAMPED_EMBEDDINGS_STREAM = 'timestamped_embeddings_out';
@@ -151,14 +147,8 @@ export class AudioEmbedder extends AudioTaskRunner<AudioEmbedderResult[]> {
protected override process(
audioData: Float32Array, sampleRate: number,
timestampMs: number): AudioEmbedderResult[] {
// Configures the number of samples in the WASM layer. We re-configure the
// number of samples and the sample rate for every frame, but ignore other
// side effects of this function (such as sending the input side packet and
// the input stream header).
this.configureAudio(
/* numChannels= */ 1, /* numSamples= */ audioData.length, sampleRate);
this.addDoubleToStream(sampleRate, SAMPLE_RATE_STREAM, timestampMs);
this.addAudioToStream(audioData, timestampMs);
this.addAudioToStreamWithShape(audioData, /* numChannels= */ 1, /* numSamples= */ audioData.length, AUDIO_STREAM, timestampMs);
this.embeddingResults = [];
this.finishProcessing();