How to generate calibration table for Yolov2 with TensorRT 7.0

andy.linluo · July 16, 2020, 4:52pm

Description

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

With Deepstream 5.0, it is possible to generate a calib table for yolov2 and run it in int8 – below link was used.

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.

I want to know how to generate the calib table for Yolov2 with TRT 7.

Environment

TensorRT Version: 7.0
GPU Type: Telsa T4
Nvidia Driver Version: 440.64.00
CUDA Version: 10.2.89
CUDNN Version:
Operating System + Version: Ubuntu 16.04.3 LTS
Python Version (if applicable):
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):

Relevant Files

Please attach or include links to any models, data, files, or scripts necessary to reproduce your issue. (Github repo, Google Drive, Dropbox, etc.)

Steps To Reproduce

Please include:

Exact steps/commands to build your repro
Exact steps/commands to run your repro
Full traceback of errors encountered

AakankshaS · July 16, 2020, 8:34pm

Hi @andy.linluo,
Please refer to the below link to check for setting up the calibrator.
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/sampleINT8#setup-the-calibrator
Also, please check if this link works for you

github.com

anguoyang/deepstream-plugins/blob/master/data/yolo/calibration/yolov2-calibration.table

1
(Unnamed ITensor* 52): 3d73b75d
(Unnamed ITensor* 44): 3dd4f487
(Unnamed ITensor* 38): 3dac6f57
(Unnamed ITensor* 49): 3d1d6d61
(Unnamed ITensor* 20): 3d457f78
(Unnamed ITensor* 35): 3e0965ed
(Unnamed ITensor* 55): 3d25fa54
(Unnamed ITensor* 41): 3dfb691e
(Unnamed ITensor* 33): 3d418ff1
(Unnamed ITensor* 30): 3d71ffd1
(Unnamed ITensor* 5): 3d333a81
(Unnamed ITensor* 2): 3c008912
(Unnamed ITensor* 54): 3dd8abc5
(Unnamed ITensor* 25): 3e0294ae
(Unnamed ITensor* 1): 4000890a
(Unnamed ITensor* 13): 3d92f46e
(Unnamed ITensor* 67): 3d186335
(Unnamed ITensor* 32): 3dcaf489
(Unnamed ITensor* 18): 3e208e5b

This file has been truncated. show original

Thanks!

andy.linluo · July 17, 2020, 7:57pm

Thanks @AakankshaS. I need Yolo2 calibration table generated with TensorRT7. The calibration table you provided was generated by older version.

AakankshaS · July 22, 2020, 8:24pm

Hi ,
You can choose to provide the implementation for writeCalibrationCache() to write calibration table
Something like this
writeCalibrationCache(&(mCalibrationCache[0]), mCalibrationCache.size())

Thanks!