uptodiff / Tensorflow-TensorRT

This repository is for my YT video series about optimizing a Tensorflow deep learning model using TensorRT. We demonstrate optimizing LeNet-like model and YOLOv3 model, and get 3.7x and 1.5x faster for the former and the latter, respectively, compared to the original models.

Don't Send Me An Email (Private Message)

Hi guys, please write your question/issue report in this repo, not by sending me a private message. So, other people can also see in case they have a similar topic, or I can refer to them.

About This Repo

This repository is for my YouTube video series here, about optimizing deep learning model using TensorRT. We demonstrate optimizing LeNet-like model and YOLOv3 model, and get 3.7x and 1.5x faster for the former and the latter, respectively, compared to the original models. For the details and how to run the code, see the video below.

List of Videos

1. Optimizing Tensorflow Model to TensorRT
2. Visualizing Before and After TensorRT Optimization
3. Optimizing Keras Model to TensorRT
4. Train Keras Model Using Your Own Dataset
5. Restore and Use Stored Keras Model to Perform Inference
6. Optimizing YOLOv3 using TensorRT
7. Another YOLOv3 Detection Result (Native Tensorflow vs TensorRT optimized)

Pre-requirement

1. TensorRT: follow the tutorial here for Ubuntu dekstop or here for Jetson devices, to install tensorRT

Library I Use:

1. Tensorflow 1.12 (Dekstop) and Tensorflow 1.11 (Jetson TX2)
2. OpenCV 3.4.5
3. Pillow 5.2.0
4. Numpy 1.15.2
5. Matplotlib 3.0.0

Environment used in this video series

Dekstop PC

1. OS: Ubuntu 16.04 - 64bit
2. GPU: GeForce 1060 6Gb
3. Driver version: 384.130
4. RAM: 16Gb
5. CUDA: 9.0
6. CuDNN: 7
7. TensorRT: 4.1.2
8. Python: 64-bit, version 3.5

Jetson TX2

1. OS: Ubuntu 16.04 - 64bit
2. GPU (with RAM shared): 8Gb
3. Driver version: Given along flashing with Jetpack 3.3
4. CUDA: 9.0
5. CuDNN: 7
6. TensorRT: 4.1.3
7. Python: 64-bit, version 3.5

Dataset

Download (subset of) MNIST dataset here, extract and put in folder dataset.

YOLOv3 Frozen Model

Download here, extract and put in folder model/YOLOv3

License: The MIT License

Copyright 2019 - Ardian Umam

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, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.