-
Hi @davidsandberg ,
I'm studying the Machine Learning on Face Recognition recently and I found that your project is a great start so I'm reading all the python codes recently.
However, I found …
-
after past 4 hours, I'm writing this issue,
when I hit npm install package, installation process stuck in this line and does not past after 4 hours, through this process my CPU worked in 100% power a…
-
When managed websockets are enabled we use array pooling to cut down on constant managed heap allocations for the byte arrays that store the sync messages Core wants us to send. However there is a cha…
-
can use this to get the shard and Region
`https://127.0.0.1:(lockfile::port)/rnet-product-registry/v4/patch-states/products/valorant/patchlines/live + Basic riot:(lockafile::password)`
will retur…
-
您好,我最近在尝试使用keras和tensorflow复现mtcnn。
有些地方我是按照自己的理解写的,训练PNet时最终的召回率只有73%左右,准确率89%左右,
我这里分类的loss在0.26左右波动,框回归loss在0.016,关键点回归在0.021左右。这样的结果是不是太糟糕了呢?
我训练的时候是实时读数据并做数据增强,一个batch里面各样本数量是30,10,10,20。
但我之…
-
楼主,你好,当我重新按照你给的方式训练完模型(数据集采用WIDER_train),发现误检率还是挺高的,同时,人脸检测过程耗时严重(图像:1262*750)约3.6s,请问楼主有相同的困扰吗?或者可以指明一下优化思路,感谢!
-
Hi Seanlinx,
I followed your instruction to generate training data for the 3 CNN networks (P, R, O_Net) from WIDER_FACE training dataset. Everything seems to work well since no error appears.
Ho…
-
gen_12net_data.py 文件中 “if np.max(Iou) < 0.3:” 就是负样本,但是这个判断某些情况下我认为是有问题的,例如随机生成的crop_box的面积大于最大单个人脸面积的3.3倍时,这个判断条件会满足,我认为这个这个负样本包含有完整的人脸,不是有效的负样本。你怎么看待这种情况?我觉得要解决这个问题,应该限制crop_box的面积大小或者判断overlap部分与人脸b…
-
Heads-up @nikolai-b, the new method of generating raster tiles works 1000+ times faster and seems to yield good results. See image below for London and the data here for the census scenario: https://g…
-
I've come across an issue while transforming large amounts of data (in volume and weight) to a Dataframe while using `REngine.CreateDataFrame` on a console application. It works well when the dataset …