Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.
6478 Discussions

Problem with INT8 Quantization

Di_Fiore__Cristian
Beginner
‎07-11-2020 03:03 AM
517 Views

Hi, I am trying to optimize my FP32 model in the INT8 model using the workbench.
At the end of the optimization, I use the INT8 model in my application, but the results are not the same as the previous model(Wrong results).
Is it due to the validation set?

I am using the latest version of OpenVINO.

Best regards, Cristian.

0 Kudos

Link Copied

1 Reply
Max_L_Intel
Moderator
‎07-13-2020 11:08 AM
485 Views

Hi Cristian @Di_Fiore__Cristian 

Please share more details about your case - what's the model, application (sample/demo), calibration method (maximum performance calibration, maximum accuracy calibration), etc.

Depending on calibration method you used, during INT8 quantization you should expect some accuracy drop as a sacrifice for performance improvement. Please see more details here https://docs.openvinotoolkit.org/latest/_docs_Workbench_DG_Int_8_Quantization.html

Best regards, Max.

0 Kudos
Copy link
Reply

Community support is provided Monday to Friday. Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.