|
MayaFlux 1.0.0
Digital-First Multimedia Processing Framework
|
Analysis result of a PhonemeAnalyzer. More...
#include <PhonemeAnalyzer.hpp>
Public Member Functions | |
| std::shared_ptr< Kakshya::TensorContainer > | embedding_tensor (std::string name) const |
| The states as a frames by hidden_width float32 tensor. | |
| std::shared_ptr< Kakshya::TensorContainer > | posterior_tensor (std::string name) const |
| The probabilities as a frames by classes float32 tensor. | |
Public Attributes | |
| std::vector< uint32_t > | best |
| Most probable class of each frame. | |
| PhonemeAnalyzerConfig | config |
| Settings the analysis was produced under. | |
| std::vector< float > | embeddings |
| State of every frame before the final layer, HubertPhonemeModel::hidden_width values each. | |
| std::vector< PhonemeEvent > | events |
| The phonemes in order: every frame whose most probable class is a phoneme and differs from the frame before. | |
| std::vector< std::pair< size_t, size_t > > | frame_positions |
| First and one past the last sample of each frame. | |
| size_t | frames { 0 } |
| Frames of the utterance. | |
| std::vector< float > | posteriors |
| Probability of every class on every frame, frame after frame, HubertPhonemeModel::classes values each. | |
Static Private Member Functions | |
| static std::shared_ptr< Kakshya::TensorContainer > | as_tensor (const std::vector< float > &values, uint64_t width, std::string name) |
Analysis result of a PhonemeAnalyzer.
Definition at line 89 of file PhonemeAnalyzer.hpp.