CIIL Anuvani-LID

bd48fd1 6 days ago

1.25 kB

// Copyright (c) 2026, CIIL Anuvani-LID. Apache-2.0.
// The sample wrapper android/LanguageIdentifier.kt in ART (tests/android/run_on_emulator.sh): identify() on a long
// recording and classify() on clips, printed as the JSON lines of tests/bindings.
package com.example.anuvanilid

import com.k2fsa.sherpa.onnx.WaveReader
import java.io.File

fun main(args: Array<String>) {
    val modelDir = File(args[0])
    val accelerator = Accelerator.valueOf(args[1])
    val longWav = args[2]
    val work = File(modelDir, "work").apply { mkdirs() }
    LanguageIdentifier(modelDir, work, accelerator, threads = 4).use { lid ->
        for (path in args.drop(3)) {
            val w = WaveReader.readWaveFromFile(path)
            val ev = lid.classify(w.samples, w.sampleRate).joinToString(",") { "[\"${it.language}\",-1,${it.probability}]" }
            println("{\"file\":\"${File(path).name}\",\"events\":[$ev]}")
        }
        val w = WaveReader.readWaveFromFile(longWav)
        val r = lid.identify(w.samples, w.sampleRate)
        println("{\"pipeline_file\":\"${File(longWav).name}\",\"pipeline\":${r.json}}")
        System.err.println("wrapper: ${r.language} ${r.confidence} dominant ${r.dominantByTime} ${r.segments.size} segments")
    }
}

	// Copyright (c) 2026, CIIL Anuvani-LID. Apache-2.0.
	// The sample wrapper android/LanguageIdentifier.kt in ART (tests/android/run_on_emulator.sh): identify() on a long
	// recording and classify() on clips, printed as the JSON lines of tests/bindings.
	package com.example.anuvanilid

	import com.k2fsa.sherpa.onnx.WaveReader
	import java.io.File

	fun main(args: Array<String>) {
	val modelDir = File(args[0])
	val accelerator = Accelerator.valueOf(args[1])
	val longWav = args[2]
	val work = File(modelDir, "work").apply { mkdirs() }
	LanguageIdentifier(modelDir, work, accelerator, threads = 4).use { lid ->
	for (path in args.drop(3)) {
	val w = WaveReader.readWaveFromFile(path)
	val ev = lid.classify(w.samples, w.sampleRate).joinToString(",") { "[\"${it.language}\",-1,${it.probability}]" }
	println("{\"file\":\"${File(path).name}\",\"events\":[$ev]}")
	}
	val w = WaveReader.readWaveFromFile(longWav)
	val r = lid.identify(w.samples, w.sampleRate)
	println("{\"pipeline_file\":\"${File(longWav).name}\",\"pipeline\":${r.json}}")
	System.err.println("wrapper: ${r.language} ${r.confidence} dominant ${r.dominantByTime} ${r.segments.size} segments")
	}
	}