Bureaucrats, dave_user, Administrators
4,650
edits
Changes
→Results
|[[File:ML - TF1.15QAT fruitsmodel.png|none|thumb|1000x1000px]]
|}
The following images show the graphs of the models after conversion (click to enlarge):
* ~ 330 ms with the half-quantized model;
* ~ 200 ms with the fully-quantized model.
{| class="wikitable"
|+
!Model
!Threads parameter
!Inference time
[ms]
|-
|'''Floating-point'''
|unspecified
|220
|-
|'''Half-quantized'''
|
|330
|-
| rowspan="2" |'''Fully-quantized'''
|unspecified
|200
|-
|4
|
|}
The total prediction time takes into account the time needed to fill the input tensor with the image and the average inference time over three predictions.
The same tests were repeated using an ext4 file system stored on a microSD card. No significant variations in the prediction times were observed.
