Use yolov3 to train yourself to make datasets and get started quickly

In object detection and classification ,Yolo It can solve many problems quickly and well , Here's a quick start Yolov3 Methods , Directly and quickly train your own data set to use .

I provide a source package that I have debugged , Contains data sets and source code , Learners can download it first and then learn it together , My next explanation is , Will be based on this source package to explain , The source package download link address is ： Add link description The extraction code is ：6vxv
Download the unzipped file sample paper as shown below ：

Let's start by introducing how to use the source code package to train your own data sets ：

1.1 Location of image data sets and labels , See below ：



1.2 among JPEGImages The sample paper in the document is shown below ：

1.3 among Annotations The sample paper in the document is shown below ：

1.3.1 Every .xml See the following for the contents of the document ：

When training your data set , You just need to copy your own dataset images to a folder JPEGImages in , The label file is copied to the file Annotations It's all right , You don't need to rename the folder yourself , Just use the framework I gave you .

2 Making dataset labels ：
About making VOC Data sets ,yolo Data sets The detailed method of , Please refer to my other blog , link ： Add link description
Make COCO Data sets See my other blog for detailed methods of , link ： Add link description

3.1 In the folder model_data In file cls_classes.txt Write the class name when labeling in the file , See below ：

3.2 Folder model_data In file yolo_anchors.txt file , Here we mainly introduce the contents of the file , Learners do not have to modify , Keep the original default , See below ：

3.3 modify voc_annotion.py In file classes_path The path of ：

3.4 function voc_anntion.py The file will generate 6 Training to be used .txt file ,6 individual .txt See the following documents respectively ：

3.5 Modify training files train.py Medium classes_path, See below ：

3.6 Direct operation train.py The file is ready for training , See below ：

4.1 Test the model after training , Copy the trained model to yolo.py Under the document , And modify it classes_path, See below ：

4.2 Start to verify the detection effect of the model after training , Run files directly predict.py file , See below ：

4.3.1 See the following table for the output after operation ：

4.3.2 See the following for test results ：

4.4.1 The code modification when you want to use video detection is shown below :

4.4.2 See the following table for the real-time video detection effect （ Here is just one of the frames , Running the code video can effectively detect the face in real time ）：

The above is the use of Yolov3 Training your own data sets , A quick way to get started , Learners only need to follow the above steps when using , You can train your own data set by modifying a few file parameters , I hope you are learning Yolov3 You have helped , Want to learn quickly Yolov5 Of scholars , See my other blog , Support a lot , thank you ！