Giter Club home page Giter Club logo

Comments (38)

MirrorYuAI avatar MirrorYuAI commented on June 8, 2024

这个106点的landmark效果好像不是很好,我这边没有匹配的人脸检测算法,直接用标注的rect框提取。结果如下
image

原图如下:
image_0003

过几天我加一个MTCNN的检测接口你再试试~

from ncnn_example.

zhanglaplace avatar zhanglaplace commented on June 8, 2024

@1976277169 好的,多谢~ 。

from ncnn_example.

Tomhouxin avatar Tomhouxin commented on June 8, 2024

我也想尝试一下,把左博的那个mtcnn加进来试试

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

来来来,可以把左博所有的模型转一波,再出一个项目~

from ncnn_example.

Tomhouxin avatar Tomhouxin commented on June 8, 2024

要跟着你学学,哈哈,我知识层面还太窄了

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

要跟着你学学,哈哈,我知识层面还太窄了

跟左博学习吧,认认真真的啃ZQCNN~

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

可以试一下重新上传的大模型,可能效果会好一点~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

可以试一下重新上传的大模型,可能效果会好一点~

能否把输入图像尺寸变大一点?比如,224x224, 或416x416, 效果应该会好一点?

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

可以试一下重新上传的大模型,可能效果会好一点~

能否把输入图像尺寸变大一点?比如,224x224, 或416x416, 效果应该会好一点?

网络输入是固定的,如果修改网络输入大小,需要修改模型结构,最后很可能不能实时~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

227f1a934eae8647e87ac09742fd041f
96x96的效果, 大佬,我这效果也太差了, 你测试的有这么差吗?

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

可以将你的测试代码上传一下吗?我这边测试没这么渣,这个模型应该是比之前上传的模型要好的~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

代码:
https://github.com/taylorguo/MTCNN_Landmark106

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

ncnn::Mat in = ncnn::Mat::from_pixels_resize((const unsigned char*)indata, ncnn::Mat::PIXEL_BGR, width, height, 48, 48);
LOGI(" ******** JNI 图像格式转换完成, Mat已读入, 大小已经更改为(96x96).");

这个是你最终的测试代码吗?
你需要将这里的48,48改成96,96~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

用的96的
667f7779706e734b7cdad4204c55f814

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

java代码看着有点晕,传入关键点检测的image需要是对应人脸区域,也就是说对人脸检测结果进行关键点定位,你这里关键点检测传入的是人脸检测结果吗?

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

是的,我用https://github.com/moli232777144/mtcnn_ncnn
把人脸方框提取出来, 然后把方框裁剪出来放到网络里面推断

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

你将人脸检测框也画上去看看~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

63aea9727648d12fa8a6a759673c4a47

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

48x48的模型测试过了吗?

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

测过了,上面发的第一个测试图是48X48的,下面这个是96x96的
难道是因为图像通道数有问题?
Android里面传进来的是4通道的

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

你的测试结果发一下看看呗?

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

给我一张你的测试图片,我晚上测试一下给你结果~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

e9775c1a76039938cbe9fd252980b9d3

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

应该是from_pixels_resize()第二个参数出问题了,改成PIXEL_RGBA2BGR,稍微好一点,但还是有问题
8e7f9208d06cf033fab55e713bbfcaf4

你帮我看看怎么改?
我还是要仔细看看ncnn源代码

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

那就是你代码有问题咯,你仔细对着改吧,我这边模型肯定是没问题的,都测试过的~

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

这个参数应该是PIXEL_RGB2BGR

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

你可以参考下我给的测试代码~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

你的测试效果,你还是帮我测一下,发我一下,我看看我能优化到什么程度

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

不好意思,不会java,这个我心有余而力不足我最多在linux或者windows上测试一下

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

是的,你在其他系统上测试的结果

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

图像的通道已经改成3通道了,还是有问题,检测出来的landmark分布和上面这张图一样

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

我也是有点晕了,难道非要导个opencv的包进去??

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

image
你自己代码写的有问题,好好检查一下吧~

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

模型出来的点坐标没问题,Java画点有问题.

from ncnn_example.

taylorguo avatar taylorguo commented on June 8, 2024

推断的时候,能否改成接受灰度图?

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

不接受,你重新提个issue吧,这个已经关闭了,我新加了retinaface做人脸检测,可以去测试一下~

from ncnn_example.

deep-practice avatar deep-practice commented on June 8, 2024

retinaface你是转的哪个项目的

from ncnn_example.

MirrorYuChen avatar MirrorYuChen commented on June 8, 2024

有问题,你可以重新开一个issue,找半天才找到retinaface转的是insightface的吧,忘记了,很久以前的事了

from ncnn_example.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.