-
你好,我在执行第9步是出现了这个错误,请问是什么原因?
-
我对PNET做了一些改动,训练出来的PNET准确率95.3%. 可是用95.3%的PNET模型生成的RNET数量里,正样本数量只有可怜的200多个,负样本有几百万个。 是我遗漏了什么重要的内容吗?
-
After upgrade to RTX 2070, I get following error in extract (mtcnn)
The Others (dlib) work.
2018-11-14 01:15:51.906117: E C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\36\tensorflow\stream_exec…
g0147 updated
5 years ago
-
>>>>>> Start tfrecord create...Stage: rnet
Converting[rnet]: 138401/138483Segmentation fault
具体原因是什么呢?
cnzdc updated
6 years ago
-
A while back an [advance way to describe the configuration variables](https://developers.home-assistant.io/docs/en/documentation_create_page.html#configuration) was introduced.
```yaml
{% configu…
-
Dear Lin,
Thank your for your great work. It is very helpful.
I have truble in preparing training data for R-Net training. the usage of gen_hard_example.py.
The code in line 153: imdb = IMDB("w…
-
首先,非常感谢你分享的源代码,给我们带来了很大的帮助。 我在gen_hard_example.py里发现必须调用你提供的两个预训练模型,否则后面会出问题。 我很困惑,你在训练出这两个预训练模型前,你肯定没有这两个模型的,那你是怎么继续后面的过程?
-
我这边测出来的为92%左右,pnet为98%,相对来看rnet下降很多,请问这是否正常,你测出来的为多少?
-
Ubuntu 16.04
Python 3.6
tensorflow (1.8.0)
tensorflow-serving-api (1.9.0)
===
I used tf.keras.layers to build pnet/rnet/onet in MTCNN, instead of writing my own layers like what the authors did. …
-
我用caffe训练的,暂时没有关键点任务,也没ohme.训练比例1:1:3(pos,part,neg),训练后,三个网络的准确率在92 93 95 左右,网络收敛。
问题
1 测试发现最后误检较多。
2 用pnet生成hard样本的时候,发现只能有2w左右的pos生成(与你的14w,差异很大);于是,我在训练rnet的时候是按照类似前面直接随机裁剪的方式生成的正样本,负样本是用pnet检测生…