train with own data/objects about pytorch-yolov3 HOT 6 CLOSED

Hello,
I realised how to make it work with my own dataset. I modifed the .cfg file, the .names file and .data and it worked. I followed the instructions of the repo that I mentioned above.

from pytorch-yolov3.

@lupotto Were you able to get results during prediction?
if yes can you please share what changes you have made to get it working?
#34 I am facing this problem.
would really appreciate your help

from pytorch-yolov3.

Resolved by uncommenting this line
From here
model = Darknet(opt.model_config_path)
#model.load_weights(opt.weights_path)
model.apply(weights_init_normal)

To here

model = Darknet(opt.model_config_path)
model.load_weights(opt.weights_path)
#model.apply(weights_init_normal)

from pytorch-yolov3.

Hi @0merjavaid,

I tried to to make it work with my own dataset/labels, but I thought that the training went smoothly and when I tried to do the testing & prediction didn't work (mAP 0 & 0 detections). Also I tried to retrain with coco dataset doing the same process and happened the same.

Did you resolve the issue changing this line of train.py? After that you loaded the new weights of your new objects/datasets and worked?

Thank you!

from pytorch-yolov3.

With your change worked if I train in coco dataset, but with my own dataset and labels doesn't. If I find a solution I would let you know

from pytorch-yolov3.

Hi, let's keep training issues in #2.

from pytorch-yolov3.