Quantization parameters that corresponds to the table, QuantizationParameters
, in the
TFLite
Model schema file.
Since per-channel quantization does not apply to input and output tensors, scale
and
zero_point
are both single values instead of arrays.
For tensor that are not quantized, the values of scale and zero_point are both 0.
Given a quantized value q, the corresponding float value f should be:
f = scale * (q - zero_point)
QuantizationParams(float scale, int zeroPoint)
|
float |
getScale()
Returns the scale value.
|
int |
getZeroPoint()
Returns the zero point value.
|
Creates a MetadataExtractor.QuantizationParams
with scale
and zero_point
.
scale | The scale value used in quantization. |
---|---|
zeroPoint | The zero point value used in quantization. |
Returns the scale value.
Returns the zero point value.