Jack Yu

@iperov The background can be removed by face landmark. No Few Shot methods like DeepFake may only help little in some instant environment. Some application scenarios that have no time...

Are you stupid? Have you anything to show? Is that the one modified from deepfake and used it to fool 10k stars?

https://github.com/szad670401/oinbagCrawler 写了个爬虫 希望对大家有帮助

cnn很慢的。就算fast_rcnn 用cpu也是到不了实时的，我试过。而且效果不好。hough line鲁邦性也估计不是很好。

自己训练的

fast-rcnn 不过我估计差不了多少..而且精度也远没有原来基于形态学的高。

it only support keras 1.x with theano data_order