Error using trainYOLOv​4ObjectDet​ector>iPar​seInputsYO​LOv4

When I wanted to train YOLOv4 to detect two classes instead of one, I changed

classes = 'class1';

to

classes = {'class1','class2'};

.

I got the following message.

Error using trainYOLOv4ObjectDetector>iParseInputsYOLOv4
The class names specified in the detector must match the class names in training data.

Error in trainYOLOv4ObjectDetector (line 114)
[trainingData, params] = iParseInputsYOLOv4(trainingData,detector,options,mfilename,varargin{:});

I found that there was an issue on line 469 in the trainYOLOv4ObjectDetector.m file.

if ~all(strcmp(gTruthClassNames',detectorClassNames))

I went to check these two variables (gTruthClassNames and detectorClassNames).

I found that one shape is 1*2 and the other is 2*1.

I'm not sure if this is the problem.

Is there an example of training YOLOv4 with more than two classes?

I know this webpage (yolov4objectdetector) describes how to use two classes in YOLOv4ObjectDetector, but the issue with my program is in trainYOLOv4ObjectDetector.

Answer 1

[Shuba Nandini]

Hello,

As per my understanding the error is due to mismatch between the names of the classes you have defined and the names of the classes in the “trainingData” being passed to the trainYolov4Detector. The issue is not with the sizes of the “gTruthClassNames” and “detectorClassNames”, as transpose has been applied, in the “strcmp” function.

You can also refer to this link for getting an idea on Multiobjectdetection: Multiclass Object Detection Using YOLO v2 Deep Learning - MATLAB & Simulink

I hope this helps!

[JENHAO HSU]

Thank you! I have solved this problem! It is caused by the different names of the classes .