PFLD-pytorch

PFLD-pytorch copied to clipboard

模型精度问题

8

感谢你的分享和工作！ 问题：我自己训练出来的基于ALFW98点数据集的模型相比你给的模型，错误率高了10个点，不知道是哪里有问题，万分感谢。

hejing-maker

PFLD 0.25X implementation

1

I'd like to try out PFLD 0.25X performance, but it's not provided. I made some changes on `__init__()` of `pfld.py` as follows: ``` class PFLDInference(nn.Module): def __init__(self, width_mult=1.0): super(PFLDInference, self).__init__()...

in-die-nibelungen

Hello, Thank you for your nice repository I make some tests for face with mask but the model cannot detect landmarks because face detection is not generated any bouding box....

hathubkhn

Does models also provides each landmark confidence score? if no then have any idea how we can get from network?

sainisanjay

Traceback (most recent call last): File "E:\code\PFLD-pytorch-master\train.py", line 221, in main(args) File "E:\code\PFLD-pytorch-master\train.py", line 146, in main optimizer, epoch) File "E:\code\PFLD-pytorch-master\train.py", line 61, in train features, landmarks = pfld_backbone(img) File...

aprillysea

可以知道每个点对应的类型吗？比如眼睛、眉毛、鼻子或是脸部轮廓。

1

maopian2333

Thank you for your great repository! Can you add a LICENSE file to specify the license of the source code and the pretrained models? I want to use it in...

vietanhdev

再将onnx模型转为动态批次后进行批量推理时报错，查了一下说是gemm_106算子维度不匹配，以batch_size=8为例，我这里的输入是{1024, 32}，而你的weight是{32，128}： 

yfding98