ForestWang comments

Results 16 comments of


                                            ForestWang

the test performance is worse than your's

![predictions](https://user-images.githubusercontent.com/5412083/87397809-7f2d1680-c5e7-11ea-8698-fa897597deed.jpg)

how do you get mobilefacenet feature?

Hi honghu: For the code "getFeatByMobileFaceNetNCNN" you replyed, is there some preprocess, i mean substract the mean value and norm? thanks very much.

progressbar error

Thanks a lot.

2019年2月14号公布的MTCNN版本的模型没有用

左博士最后这个问题怎么解决的？

2019年2月14号公布的MTCNN版本的模型没有用

不行，试过了。

等新代码更新

大佬结果论文结果咋样了？

could you share the pretrained weights and which open seaship dataset you are using?

thanks! due to my dataset is very small, even though pretrained with coco of yolov5. the rotation angle theta is very bad. so i want to use your weights as...

最大视差参数对算法的影响是什么？

请教个基础问题，再做深度匹配之前，左右图是不是应该1. 先去畸变，2. 双目校正，然后再做匹配？

How to convert yolov5 to NCNN?

> refer https://zhuanlan.zhihu.com/p/275989233 (in chinese) many thanks!

How to train the models

all loss will be nan after about 15 epoch, anyone knows why?