https://github.com/walktree/libtorch-yolov3

A Libtorch implementation of the YOLO v3 object detection algorithm
https://github.com/walktree/libtorch-yolov3

cpp libtorch pytorch yolov3

Last synced: 3 months ago
JSON representation

A Libtorch implementation of the YOLO v3 object detection algorithm

• Host: GitHub
• URL: https://github.com/walktree/libtorch-yolov3
• Owner: walktree
• Created: 2018-11-28T08:12:01.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-10-16T03:51:34.000Z (over 4 years ago)
• Last Synced: 2024-08-03T01:23:33.697Z (6 months ago)
• Topics: cpp, libtorch, pytorch, yolov3
• Language: C++
• Size: 1.54 MB
• Stars: 449
• Watchers: 15
• Forks: 130
• Open Issues: 37
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-yolo-object-detection - walktree/libtorch-yolov3 - yolov3?style=social"/> : A Libtorch implementation of the YOLO v3 object detection algorithm, written with pure C++. (Other Versions of YOLO)
• awesome-yolo-object-detection - walktree/libtorch-yolov3 - yolov3?style=social"/> : A Libtorch implementation of the YOLO v3 object detection algorithm, written with pure C++. (Other Versions of YOLO)

README

        
# libtorch-yolov3

A Libtorch implementation of the YOLO v3 object detection algorithm, written with pure C++. It's fast, easy to be integrated to your production, and CPU and GPU are both supported. Enjoy ~

This project is inspired by the [pytorch version](https://github.com/ayooshkathuria/pytorch-yolo-v3), I rewritten it with C++.

## Requirements

1. LibTorch v1.0.0

2. Cuda

3. OpenCV (just used in the example)

## To compile

1. cmake3

2. gcc 5.4 +

```

mkdir build && cd build

cmake3 -DCMAKE_PREFIX_PATH="your libtorch path" ..

# if there are multi versions of gcc, then tell cmake which one your want to use, e.g.:

cmake3 -DCMAKE_PREFIX_PATH="your libtorch path" -DCMAKE_C_COMPILER=/usr/local/bin/gcc -DCMAKE_CXX_COMPILER=/usr/local/bin/g++ ..

```

## Running the detector

The first thing you need to do is to get the weights file for v3:

```

cd models

wget https://pjreddie.com/media/files/yolov3.weights 

```

On Single image:

```

./yolo-app ../imgs/person.jpg

```

As I tested, it will take 25 ms on GPU ( 1080 ti ). please run inference job more than once, and calculate the average cost.