Closed himikox closed 4 years ago
@himikox you likely have environment issues. I'll link you to our working environments you can use for training here:
To access an up-to-date working environment (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled), consider a:
i'm already using google colab notebook, is it not enough?
@himikox ah, yes then you should be all set on your environment. Everything you posted seems fine, though your model does not seem to be based on the default yolov3-spp.cfg, which should say roughly:
Model Summary: 225 layers, 6.29987e+07 parameters, 6.29987e+07 gradients
Perhaps that is the cause?
@himikox it looks like you may have used yolov3.cfg. Probably best to stick with the defaults, as remember if you try to load --weights that are incompatible with your cfg you'll probably get errors. If in doubt, you can load no weights using --weights ''
thanks a lot i'll do as you said 💯
@glenn-jocher Hello , training works fine now but when i'm trying to detect an image no boxes are shown . what should i do?
@himikox you're going to need to supply a hell of a lot more information than that if you expect anyone to be able to help you.
Hello. i'm using visual genome as my custom dataset. i followed all the steps as it says in here https://github.com/ultralytics/yolov3/wiki/Train-Custom-Data
all requirements already installed
i have the images and labels folder : (i am using 200 images)
example of the content of a label file :
this is the content of my visualgenome_train.txt file (in /data)
my visualgenome.names file contains 42 class names
and my visualgenome.data looks like this :
also i modified the 3 yolo layers in yolov3.cfg file to look like this :
when i train this happens :
or it just crashes without displaying any error message.
i'm confused right now idk if i'm missing something.
EDIT : i got rid of empty lines in some files i thought maybe that could be the reason but i'm still getting this :