PyTorch-YOLOv3

A minimal PyTorch implementation of YOLOv3, with support for training, inference and evaluation.

Installation

Clone and install requirements

$ git clone https://github.com/wzj5133329/pytorch-yolov3
$ cd pytorch-yolov3/
$ sudo pip3 install -r requirements.txt

Download pretrained weights

$ cd weights/
$ bash download_weights.sh

Download COCO

$ cd data/
$ bash get_coco_dataset.sh

Train on Custom Dataset

Custom model

生成符合自有数据集的模型文件

修改类别数目

$ cd config/                                # Navigate to config dir
$ bash create_custom_model.sh <num-classes> # Will create custom model 'yolov3-custom.cfg' #修改网络中的训练类数目

修改anchors

教程

Classes

Add class names to data/custom/classes.names. This file should have one row per class name.

修改 config/custom.data文件

Image Folder

Move the images of your dataset to data/custom/images/.

Annotation Folder

Move your annotations to data/custom/labels/ (与imgaes文件夹匹配，且label要与classes.names顺序匹配) Each row in the annotation file should define one bounding box, using the syntax label_idx x_center y_center width height. The coordinates should be scaled [0, 1], and the label_idxshould be zero-indexed and correspond to the row number of the class name indata/custom/classes.names`.

Define Train and Validation Sets

生成包含训练与验证集所有图片路径的 trian.txt与valid.txt文件 使用 data/deal/imagename2txt.py 文件 In data/custom/train.txt and data/custom/valid.txt, add paths to images that will be used as train and validation data respectively.

Train

To train on the custom dataset run:

$ python3 train.py --model_def config/yolov3-custom.cfg --data_config config/custom.data

Add --pretrained_weights weights/darknet53.conv.74 to train using a backend pretrained on ImageNet.

Demo

To test on one image :(ESC退出) (需要安装opencv)

$ pip install opencv-python

$ python3 demo.py --image_path=./data/samples/dog.jpg

To test on folder & save the images :

$ python3 detect.py --image_folder data/samples/

[Paper] [Project Webpage] [Authors' Implementation]

参考 https://github.com/eriklindernoren/PyTorch-YOLOv3