Training my YOLOv2 object detector failed

3 views (last 30 days)
Jesús Silva
Jesús Silva on 2 May 2022
Commented: Jesús Silva on 8 May 2022
I have created my network using the pre-trained network resnet18, and with the following code:
numClases = 9;
tamanio = [224 224 3];
anchorBoxes = estimateAnchorBoxes(TrainingEscalado,numClases)
red = resnet18();
arquitectura = yolov2Layers(tamanio,numClases,anchorBoxes,red,"res5b_relu","ReorgLayerSource","res3a_relu")
"TrainingEscalado" is the datastore for training already scaled to network's inputs size. When I import the network into the Deep Network Designer app, I don't have errors, and the same happens when I import the datastore for training.
But as I click train, an error message is displayed with the following content:
And I don't know what to do to fix it, the network already has 9 classes as output so I don't know what fails.

Answers (1)

yanqi liu
yanqi liu on 7 May 2022
yes,sir,may be use yolo v3、v4 to test,such as
https://ww2.mathworks.cn/matlabcentral/fileexchange/72520-yolov3-yolov4-matlab
  1 Comment
Jesús Silva
Jesús Silva on 8 May 2022
I don't understand, the problem is the version of YOLO that I'm using? If I use YOLO v3 or v4 it would be fine?

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!