search

rpng / calc

Convolutional Autoencoder for Loop Closure
BSD 3-Clause "New" or "Revised" License
190 stars 45 forks source link

What is the value of loss? #26

Closed wangtiandong-roborock closed 3 years ago

wangtiandong-roborock commented 4 years ago

Hi @nmerrill67 , I'm finetuing calc network on our own dataset. Could you provide the final value of your training loss? Mine is 63, which I think is too large, and the PR Curve is also worse than your weights. I trained the network 500k iterations.

shenyanqing1105 commented 3 years ago

Hi @nmerrill67 , I'm finetuing calc network on our own dataset. Could you provide the final value of your training loss? Mine is 63, which I think is too large, and the PR Curve is also worse than your weights. I trained the network 500k iterations.

Excuse me, have you solved this problem? I have a similar problem now

wangtiandong-roborock commented 3 years ago

Hi @nmerrill67 , I'm finetuing calc network on our own dataset. Could you provide the final value of your training loss? Mine is 63, which I think is too large, and the PR Curve is also worse than your weights. I trained the network 500k iterations.

Excuse me, have you solved this problem? I have a similar problem now

No, I haven't. I found there were lots of other methods better than calc network. I do suggest you read more recent papers.

shenyanqing1105 commented 3 years ago

Hi @nmerrill67 , I'm finetuing calc network on our own dataset. Could you provide the final value of your training loss? Mine is 63, which I think is too large, and the PR Curve is also worse than your weights. I trained the network 500k iterations.

Excuse me, have you solved this problem? I have a similar problem now

No, I haven't. I found there were lots of other methods better than calc network. I do suggest you read more recent papers.

Got it! Thanks for your suggestion~