lhwcv comments

Results 13 comments of


                                            lhwcv

Optimization of reprojection error in C++ - looking for solver

@zacniewski Maybe nlopt is another choice. I'm also working on C++ version, but blocked by 'Powell' sorver in C++ , I have tried some other optimization methods, the result is...

请问有交流群吗？

@tjadamlee 同请求，群超了200进不去，微信id: lhwcvf，辛苦加个好友拉我一下，感谢！

Always converges to nearly 22dB

@dex1990 Have you fixed your problem? I also meet the similar problem when fitting a BW filter. @mgharbi Any suggestions? Thanks a lot! ( loss got to ~ 0.003 and...

Would you be interested in converting the ncnn model of these two？

Thank you for your good advice. I 'll try this and deploy use NCNN: https://github.com/brentspell/hifi-gan-bwe Maybe next month I'll have time to do this.

有关模型部署FFT的问题

如果实际用的时候还是要加窗的，只是这个paper 他没加，这里只是为了完全还原

Print statement

Already removed, Thanks!

Is there any way to reduce the GPU memory usage and enhance the inference speed?

You can try the TensorRT version by @rhysdg , https://github.com/lhwcv/mlsd_pytorch#benchmarks

Project dependencies may have API risk issues

pull a request is pretty welcome

Loss becomes nan after certain epoch.

> @lhwcv do you have any possible solution for the problem of nans noticed here? @michelebechini Hi, Have you tried reduce the learning rate, If still NaN, report your software's...

How to train the models

Sorry for my late reply. (A little busy at work) My result is lower than in the paper, about ~2-3 percent lower than it. Maybe next week I'll push my...