search

ZENGYIMING-EAMON / CorrNet3D

CorrNet3D: Unsupervised End-to-end Learning of Dense Correspondence for 3D Point Clouds
GNU General Public License v3.0
136 stars 21 forks source link

early stop？ #13

Open balabalaboy opened 1 year ago

balabalaboy commented 1 year ago

How many epochs did you train? It seems that it takes a long time to train. It seems that the decreasing trend of loss has levelled off, and the loss value increases instead of decreasing. Then roughly how many epochs will the training stop? image image

image

By the way, is it possible to register one could point to another cloud point single time? not the total dataset, just the single one.

TAO-TAO-TAO-TAO-TAO commented 1 year ago

How many epochs did you train? It seems that it takes a long time to train. It seems that the decreasing trend of loss has levelled off, and the loss value increases instead of decreasing. Then roughly how many epochs will the training stop? image image

image

By the way, is it possible to register one could point to another cloud point single time? not the total dataset, just the single one.

I would like to ask how long it will take you to train for approximately 100 epochs. I need over half an hour to train an epoch, using 12g of 2080ti

black-prince222 commented 1 year ago

How many epochs did you train? It seems that it takes a long time to train. It seems that the decreasing trend of loss has levelled off, and the loss value increases instead of decreasing. Then roughly how many epochs will the training stop? image image

image

By the way, is it possible to register one could point to another cloud point single time? not the total dataset, just the single one.

I have the same question. my rank loss and val loss is increasing.

black-prince222 commented 1 year ago

How many epochs did you train? It seems that it takes a long time to train. It seems that the decreasing trend of loss has levelled off, and the loss value increases instead of decreasing. Then roughly how many epochs will the training stop? image image

image

By the way, is it possible to register one could point to another cloud point single time? not the total dataset, just the single one.

Hello, do you know why the loss is an uptrend? I also encountered this situation.