Function classify_runtime_audio

src/framework/modules/titanet_runtime.cpp:720–742

}

TitaNetClassifyResult classify_runtime_audio(
    TitaNetRuntime & runtime,
    const TitaNetWeights & weights,
    const std::vector<float> & normalized_classifier_weight,
    const std::vector<std::string> & labels,
    const runtime::AudioBuffer & audio) {
    const auto start = Clock::now();
    if (labels.size() != static_cast<size_t>(weights.num_classes)) {
        throw std::runtime_error("label count mismatch");
    }
    auto embedding = embed_runtime_audio(runtime, weights, audio);
    auto logits = classify_embedding(weights, normalized_classifier_weight, embedding);
    const auto best = std::max_element(logits.begin(), logits.end());
    const int index = static_cast<int>(std::distance(logits.begin(), best));
    TitaNetClassifyResult out;
    out.label = labels[static_cast<size_t>(index)];
    out.index = index;
    out.score = *best;
    out.embedding = std::move(embedding);
    out.logits = std::move(logits);
    debug::timing_log_scalar("titanet.classify_audio_ms", engine::debug::elapsed_ms(start, Clock::now()));
    return out;
}

TitaNetRuntime::TitaNetRuntime(
    std::shared_ptr<const TitaNetWeights> weights,