新的Andr​​oid面对的API限制 [英] New Android Face API limitations

查看:312
本文介绍了新的Andr​​oid面对的API限制的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!

问题描述

我一直在测试新的脸部API realesed为Android,并注意到即使有ACCURATE_MODE启用后,它并没有检测到旧FaceDetector API用于检测人脸,而且我想知道的位图编码的影响 RGB_565与生产的结果ARGB_888。

I have been testing the new Face API realesed for android, and noticed even with "ACCURATE_MODE" enabled, it doesn't detect faces that old FaceDetector API used to detect, Also i would like to know the effect of Bitmap coding "RGB_565" vs "ARGB_888" in producing the results.

推荐答案

探测器首先将图像转换为灰度,所以RGB_565与ARGB_888可能不会有所作为。

The detector first converts the image to grayscale, so "RGB_565" vs "ARGB_888" probably doesn't make a difference.

见上面的注释有关图像大小和脸相对比例​​的图像(即,未检测到较小的面):

See the comment above about image size and proportion of the face relative to the image (i.e., smaller faces not detected):

这是一个权衡,我们在简化API设置进行的。虽然算法能够检测更小的脸部,则API中图像扫描被限制面中至少10%的所述图像宽度(默认)或至少35%的图象宽度为prominentFaceOnly设置。由于检测较小的面需要更长的时间,这些默认值被选择,以提供更好的性能折衷(在精度为代价为非常小的面)。

This is a trade-off that we made in simplifying the API settings. Although the algorithm is capable of detecting much smaller faces, image scanning in the API is limited to faces that are at least 10% the image width (default) or at least 35% the image width for the "prominentFaceOnly" setting. Since detecting smaller faces takes longer, these defaults were chosen to provide a better performance trade-off (at the expense of accuracy for very small faces).

这篇关于新的Andr​​oid面对的API限制的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!

查看全文
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆