Loss value of A different than B

Sometimes you can see that the loss value of one of your training set is smaller than the other. For example Loss of A is 0.020 and Loss of B is 0.009.

These values don’t necessarily need to be equal, as long as you know you need the value of B to me smaller, and you are happy with the result (after doing some quick previews of the final output), you can go ahead and stop the training process.

If your final video is still blurry or some angles don’t display quite right, then the problem lies with the source images. You need to add more images or delete some of the bad ones, and then restart the training process.

Leave a Reply

Your email address will not be published. Required fields are marked *