| | |
|---|
| | | # MIT License (https://opensource.org/licenses/MIT) |
|---|
| | | # Modified from 3D-Speaker (https://github.com/alibaba-damo-academy/3D-Speaker) |
|---|
| | | |
|---|
| | | import umap |
|---|
| | | import scipy |
|---|
| | | import torch |
|---|
| | | import sklearn |
|---|
| | | import hdbscan |
|---|
| | | import numpy as np |
|---|
| | | |
|---|
| | | from sklearn.cluster._kmeans import k_means |
|---|
| | |
|---|
| | | self.min_samples = min_samples |
|---|
| | | self.min_cluster_size = min_cluster_size |
|---|
| | | self.metric = metric |
|---|
| | | import hdbscan |
|---|
| | | self.hdbscan = hdbscan |
|---|
| | | |
|---|
| | | def __call__(self, X): |
|---|
| | | import umap.umap_ as umap |
|---|
| | | umap_X = umap.UMAP( |
|---|
| | | n_neighbors=self.n_neighbors, |
|---|
| | | min_dist=0.0, |
|---|
| | | n_components=min(self.n_components, X.shape[0] - 2), |
|---|
| | | metric=self.metric, |
|---|
| | | ).fit_transform(X) |
|---|
| | | labels = hdbscan.HDBSCAN( |
|---|
| | | labels = self.hdbscan.HDBSCAN( |
|---|
| | | min_samples=self.min_samples, |
|---|
| | | min_cluster_size=self.min_cluster_size, |
|---|
| | | allow_single_cluster=True).fit_predict(umap_X) |
|---|
| | |
|---|
| | | if X.shape[0] < 20: |
|---|
| | | return np.zeros(X.shape[0], dtype='int') |
|---|
| | | if X.shape[0] < 2048 or k is not None: |
|---|
| | | # unexpected corner case |
|---|
| | | labels = self.spectral_cluster(X, k) |
|---|
| | | else: |
|---|
| | | labels = self.umap_hdbscan_cluster(X) |
|---|
