本文整理匯總了Python中vggish_params.QUANTIZE_MIN_VAL屬性的典型用法代碼示例。如果您正苦於以下問題：Python vggish_params.QUANTIZE_MIN_VAL屬性的具體用法？Python vggish_params.QUANTIZE_MIN_VAL怎麽用？Python vggish_params.QUANTIZE_MIN_VAL使用的例子？那麽, 這裏精選的屬性代碼示例或許可以為您提供幫助。您也可以進一步了解該屬性所在類vggish_params的用法示例。
在下文中一共展示了vggish_params.QUANTIZE_MIN_VAL屬性的1個代碼示例，這些例子默認根據受歡迎程度排序。您可以為喜歡或者感覺有用的代碼點讚，您的評價將有助於係統推薦出更棒的Python代碼示例。
示例1: postprocess
# 需要導入模塊: import vggish_params [as 別名]
# 或者: from vggish_params import QUANTIZE_MIN_VAL [as 別名]
def postprocess(self, embeddings_batch):
"""Applies postprocessing to a batch of embeddings.
Args:
embeddings_batch: An nparray of shape [batch_size, embedding_size]
containing output from the embedding layer of VGGish.
Returns:
An nparray of the same shape as the input but of type uint8,
containing the PCA-transformed and quantized version of the input.
"""
assert len(embeddings_batch.shape) == 2, (
'Expected 2-d batch, got %r' % (embeddings_batch.shape,))
assert embeddings_batch.shape[1] == vggish_params.EMBEDDING_SIZE, (
'Bad batch shape: %r' % (embeddings_batch.shape,))
# Apply PCA.
# - Embeddings come in as [batch_size, embedding_size].
# - Transpose to [embedding_size, batch_size].
# - Subtract pca_means column vector from each column.
# - Premultiply by PCA matrix of shape [output_dims, input_dims]
# where both are are equal to embedding_size in our case.
# - Transpose result back to [batch_size, embedding_size].
pca_applied = np.dot(self._pca_matrix,
(embeddings_batch.T - self._pca_means)).T
# Quantize by:
# - clipping to [min, max] range
clipped_embeddings = np.clip(
pca_applied, vggish_params.QUANTIZE_MIN_VAL,
vggish_params.QUANTIZE_MAX_VAL)
# - convert to 8-bit in range [0.0, 255.0]
quantized_embeddings = (
(clipped_embeddings - vggish_params.QUANTIZE_MIN_VAL) *
(255.0 /
(vggish_params.QUANTIZE_MAX_VAL - vggish_params.QUANTIZE_MIN_VAL)))
# - cast 8-bit float to uint8
quantized_embeddings = quantized_embeddings.astype(np.uint8)
return quantized_embeddings