Pytorch_Retinaface

Pytorch_Retinaface copied to clipboard

Dear author: Did you try ghostnet as bonebock? How the result you get？

Alex866850

Loss balancing parameters and normalization

2

Hi @biubug6, First off, thank you for your work and for sharing. I had a question about your loss definitions. The loss defined in the paper L_cls(pi, pi^*) + lamda_1...

casperthuis

Is there any mehtod for converting mobilenet0.25 to TensorRT engine?

10

Hello all. I am currently working with MobiletNet0.25 to convert the model to TensorRT. I tried conversion with TensorRT environment supported by Nvidia. However I couldn't get success. Do you...

G-Bong

 Hi, everyone When I compute the AP, the image error result will not be compute. Is it correct ? For example, the Image grouth have one face. I use...

alicera

FDDB eval_tool is 404 not found

8

README.MD Download eval_tool to evaluate the performance. https://bitbucket.org/marcopede/face-eval that is 404 not found

saladcat

PriorBox forward is slow

2

The prior box forward function is not vectorised. This is easily vectorizable as below. ``` def priorbox_forward(min_sizes, steps, clip, image_size): feature_maps = [[ceil(image_size[0] / step), ceil(image_size[1] / step)] for step...

gan3sh500

Thanks for your outstanding work! In the train.py file, I only find the classification loss, box loss and landmark loss. So where is the dense-face regression loss used in your...

cookchicken

mobilenetV1X0.25_pretrain.tar损坏

2

mobilenetV1X0.25_pretrain.tar这个压缩包损坏了，请问有没有新的可用文件

lijiahuan1009

https://github.com/biubug6/Pytorch_Retinaface/blob/master/layers/modules/multibox_loss.py#L90 Does anyone know the setting? What does it mean.

alicera

你好，我自己的数据集只有3个关键点，所以在WiderFaceDetection类中将annotations改成了11维（4个box坐标 + 6个关键点坐标 + 1个label），但在训练时总是报一些维度转换的错误，看了下代码，里面有很多地方都写死了5个关键点，请问有什么好方法只训练3个关键点吗？期待回复

ybcc2015