INT8 calibration table for Yolov2

andy.linluo · July 16, 2020, 5:10pm

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU): Telsa T4
• DeepStream Version: 5.0
• JetPack Version (valid for Jetson only)
• TensorRT Version : 7.0
• NVIDIA GPU Driver Version (valid for GPU only): 440.64.00

DS 5.0 releases the int8 calib table for yolov3 but not yolov2.

With DS4, it is possible to generate a calib table for yolov2 and run it in int8 – below link was used.

github.com

NVIDIA-AI-IOT/deepstream_reference_apps/blob/restructure/yolo/lib/calibrator.h

/**
MIT License

Copyright (c) 2018 NVIDIA CORPORATION. All rights reserved.

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

This file has been truncated. show original

But for DS5, which uses TRT 7 – there are some changes and the above generated calib table is not working.

Is there any INT8 calibration table for Yolov2 available for use?

Or how can I generate it with Tensorrt 7?

mchi · August 4, 2020, 1:35am

Hi @andy.linluo
We don’t provide INT8 calibration table for YoloV2.
Will you use YoloV2 in your product?

For now, to generate INT8 calibration, you need to add TRT INT8 calibration in /opt/nvidia/deepstream/deepstream-5.0/sources/objectDetector_Yolo/nvdsinfer_custom_impl_Yolo/yolo.cpp by referring to the calibrtion code in TRT INT8 sample - ${tensorrt_package}/samples/sampleINT8