I am training a dataset in yolov4 using the repo from AlexeyAB darknet.

In his repo, backup weights are created every so iterations but you originally train with a pretrained weights file.

I was training with the pretrained weight file with a pretty big dataset(80k images, 80 k batches).

My power disconnected on my Computer at around only 3000 iterations. I ran the command to start training again with the back up weights file generated which began from 3000 iterations again.

Now this might be a really dumb question because I am not sure if the weights are an actual backup or a new poorly trained weights file? My question is, will this affect the results because the weights file is different? Or is this still technically the original weights file with the 3000 iterations trained with the dataset and it would not be different from restarting with the original weights file used to train?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.