Object Detection through YOLOv3 using Darknet Importer in MATLAB

Hi

I am loading the cfg and weight files using darknet importer but finding difficulties to add the detection layer at the end.

In the official Documentation here (https://www.mathworks.com/matlabcentral/fileexchange/71277-deep-learning-darknet-importer?focused=098f9ec1-f049-4aef-a910-93b617a7a299&tab=function), there is no explanation about adding a detection layer. It is only for image classification.

Can anyone please help me in this regard?

Or you can tell me any other way to detect objects in matlab (pre-trained on coco)

Answer 1

[Farah Sarwar]

You can use YOLOv2 in MATLAB for object detection as well as R-CNN, Fast R-CNN and Faster R-CNN. Here is the link of help file for object detecion in MATLAB using YOLOv2: https://au.mathworks.com/help/deeplearning/examples/object-detection-using-yolo-v2.html

YOLOv3 definitely performs better than v2 but you will develop some understanding of the topic.

[Muhammad Talha]

Hi Farah Sarwar,

Thanks for your help.

The sent link has the dataset only for vehicles.

Can you please guide me how I can get or make the .mat file for my own dataset or for the famous datasets like COCO or VOC, so that I can train the dataset for YOLO v2 or Faster RCNN or Mask RCNN?

[Hayat Bouchkouk]

Hi,Muhammad Talha

I want to know if you are able to load coco dataset in matlab or not because i face the same probleme,plizz if any body can help me ,i'me really stuck

Answer 2

[cui,xingxing]

This is the yolov3 you want, but there is a problem with saving the model during training, especially the parameter saving of the bn layer should be consistent with darknet, and the labeled [x, y, w, h], instead of Normalized [center_x, center_y, w, h ]