liqi

Me too

@so-as Hello , Did you implemented landmark localization training ?

@so-as Ok, it doesn't matter. I'll try

我使用作者的code进行训练，使用20层的resnet，竟然达到了95.04%的准确率，表示震惊，这还没有用其他技巧

Thank you for your problem, but I think your conclusion is wrong, if 260ms per image, the image window will update 3.8 frame per second, that's impossible, you can use...

Hi, thanks again for your problem, I found there is something wrong in my code , I retest my program under windows, the result shows about 68ms per image, I'...

@sophieyutingye 你好，请问你是复现了这个对吧？能否告诉下你用到了哪些数据集扩增方法，非常感谢

不好意思哈，最近忙于校招，没什么时间管理，我会尽快更新code，并发布更新的版本。不便之处，烦请见谅。

1. I have updated the code again, and I think I made a mistake before. This time, I delete the incorrect scales in the initialization stage. 2. The reasons is...

You can refer to https://github.com/Seanlinx/mtcnn