MarStarck comments

Results 20 comments of


                                            MarStarck

zero loss

same problem. but if don't use cuda the loss is normal. so I think it may be environment problem. I'm using cuda8.0, pytorch 0.4.1

zero loss

I installed cuda 9.0 with cudnn 7, pytorch 0.4.0 and remake and reinstall it and it is not 0 anymore!

Can delete this code "preds = preds.squeeze(2)"?

using python 3.6, pytorch 0.4 will meet the problem, preds is 26*64 and cannot be squeezed. I delete it but the loss is always 0. @Aurora11111 did you meet the...

mbox_loss is always 0

@noranart My training loss is properly decreased but the mbox_loss jumps from 0 to 5~7. My problem is how to define the (top left, top right, bottom right, bottom left)...

Bad detect result after training

which tool did you use for label image?

There are logic error in codes.

ha, that's a permanent dispute between fundamental analysts and technical analysts.

Will it work for characters?

I have added character training data but the acc drops to about 70%

the data source is synthetic text data by [TextRecognitionDataGenerator](https://github.com/Belval/TextRecognitionDataGenerator) I believe they are quite clear without distortion. So I think the 10-layer conv model cannot describe these characters, it's under...

Will it work for characters?

@fkoorc I'm from China, my email: [email protected]