2

I want to detect only person detection by using YOLOv3. it detecting well. but it also searching and matching for other objects. That's why it takes some more time to detect person. Now I want to create my own data set based on person. I have downloaded 85 thousand cropped images from Google API http://www.rapdataset.com/

But for train these images, it need to be labeled and annotated. thousands of images, how can I do? it takes much time to annotate. Is there any way to annotate these images programmatically like in python?

azhar
  • 351
  • 3
  • 13

1 Answers1

1

You can try this one to help you, or you have the alternative to use the already annotated set of COCO, and choose persons only of it, like in this question.

AbdelAziz AbdelLatef
  • 3,650
  • 6
  • 24
  • 52
  • i have tried to change these files coco.names. now it has one class = person and in cfg file i have changed those lines as per the given topic. Modify your cfg file (e.g. yolov3.cfg), change the 3 classes on line 610, 696, 783 from 80 to 1 Change the 3 filters in cfg file on line 603, 689, 776 from 255 to 18 (derived from (classes+5)x3) but still its not working. better same result. – azhar Oct 11 '20 at 22:39
  • followed this link as you shared. https://stackoverflow.com/questions/57898577/how-to-reduce-number-of-classes-in-yolov3-files – azhar Oct 11 '20 at 22:40
  • 1
    @azhar Have you trained it with their dataset? – AbdelAziz AbdelLatef Oct 11 '20 at 22:54
  • no just i have changed those lines from config file. and removed all classed from coco classes file. than had run it. i did not followed the advance step which was motioned blow of that topic. – azhar Oct 11 '20 at 22:59
  • 1
    @azhar This will not do it, you have to train it from the beginning. – AbdelAziz AbdelLatef Oct 11 '20 at 23:00
  • ok thanks. i'll just follow the advance concept to train it. thanks for your time. – azhar Oct 11 '20 at 23:07