jianzhnie / pyramidbox_pytorch

pytorch实现的Pyramidbox 人脸检测模型， 对原来代码的部分模块进行了修改，更简洁高效

A PyTorch Implementation of PyramidBox

PyramidBox: A Context-assisted Single Shot Face Detector.

Description

I train pyramidbox with pytorch and the result approaches the original paper result,the pretrained model can be downloaded in vgg,the final model can be downloaded in Pyramidbox.the AP in WIDER FACE as following:

	Easy MAP	Medium MAP	hard MAP
origin paper	0.960	0.948	0.888
this repo	0.948	0.938	0.880

the AP in AFW,PASCAL,FDDB as following:

AFW	PASCAL	FDDB
99.65	99.02	0.983

the gap is small with origin paper,I train 120k batch_size 4 which is different from paper,which maybe cause the gap,if you have more gpu ,the final result maybe better.

Requirement

• pytorch 1.3
• opencv
• numpy
• easydict

Prepare data

1. download WIDER face dataset
2. modify data/config.py
3. python tools/prepare_wider_data.py

Train

python main.py --batch_size 4  
		--lr 5e-4

Evalution

according to yourself dataset path,modify data/config.py

1. Evaluate on AFW.

python tools/afw_test.py

2. Evaluate on FDDB

python tools/fddb_test.py

3. Evaluate on PASCAL face

python tools/pascal_test.py

4. test on WIDER FACE

python tools/wider_test.py

Demo

you can test yourself image

python demo.py

Result

References

• PyramidBox: A Context-assisted Single Shot Face Detector
• PyramidBox model
• S3FD_pytorch
• ssd.pytorch

pyramidbox_pytorch's Issues

您好，我运行demo.py的时候，显示如下，网上说batchsize太大，但是我找到不到运行demo.py调整batchsize的地方？

/home/lyl/anaconda3/envs/py1.3/bin/python /home/lyl/000_Code/pyramidbox_1.3/demo.py
Traceback (most recent call last):
File "/home/lyl/000_Code/pyramidbox_1.3/demo.py", line 113, in
detect(net, path, args.thresh)
File "/home/lyl/000_Code/pyramidbox_1.3/demo.py", line 73, in detect
y = net(x)
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/lyl/000_Code/pyramidbox_1.3/models/pyramidbox.py", line 173, in forward
ssh_conv3_norm = self.cpm0
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/lyl/000_Code/pyramidbox_1.3/models/pyramidbox.py", line 56, in forward
out_residual = self.branch1(x)
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/lyl/000_Code/pyramidbox_1.3/models/pyramidbox.py", line 35, in forward
return self.bn1(x)
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/modules/batchnorm.py", line 81, in forward
exponential_average_factor, self.eps)
File "/home/lyl/anaconda3/envs/py1.3/lib/python3.6/site-packages/torch/nn/functional.py", line 1670, in batch_norm
training, momentum, eps, torch.backends.cudnn.enabled
RuntimeError: CUDA out of memory. Tried to allocate 322.00 MiB (GPU 0; 3.95 GiB total capacity; 2.90 GiB already allocated; 88.56 MiB free; 39.48 MiB cached)

Process finished with exit code 1

bug in demo.py

from pyramidbox import build_net
to
from models.pyramidbox import build_net

bug in demo.py

the flowing code is in your 'models.pyramidbox.py':

if not self.is_infer:
    output = (face_mbox_loc, face_mbox_conf,
                      head_mbox_loc, head_mbox_conf)
 else:
    output = self.detect(face_mbox_loc, self.softmax(face_mbox_conf))

but the function self.detect() needs three parameters in your script 'layers.functions.detection.py'

class Detect(Function):
    def __init__(self, cfg):
          ...
    def forward(self, loc_data, conf_data, prior_data):

so, when i run 'demo.py', the flowing error reported:

TypeError: forward() missing 1 required positional argument: 'prior_data'

运行demo.py时出错

你好，我成功的训练出了模型。但是在运行demo.py的时候出现了一个bug。

Traceback (most recent call last):
File "demo.py", line 117, in
detect(net, path, args.thresh)
File "demo.py", line 73, in detect
y = net(x)
File "/home/huyutao/anaconda3/envs/pytorch_gpu/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/huyutao/anaconda3/envs/pytorch_gpu/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 150, in forward
return self.module(*inputs[0], **kwargs[0])
File "/home/huyutao/anaconda3/envs/pytorch_gpu/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/huyutao/face_mask_detect/pyramidbox_pytorch/models/pyramidbox.py", line 262, in forward
self.softmax(face_mbox_conf))
TypeError: forward() missing 1 required positional argument: 'prior_data'

我查看了下源代码，是在pyramidbox.py中的261行。output = self.detect(face_mbox_loc, self.softmax(face_mbox_conf))少传了一个参数
Detect class貌似需要3个参数：loc_data, conf_data, prior_data；
请问下，prior_data这个参数如何传递？

如果我的理解有误，也请告知解决这个bug的方法。。。谢谢。

bug in train.py

criterion1(out, face_targets)
but criterion1 = MultiBoxLoss(cfg, args.cuda)
def forward(self, predictions, priors, targets):
"""Multibox Loss
need 3 args
TypeError: forward() missing 1 required positional argument: 'targets'

Problems in training

Hello, I have the following error when running the train.py file:
runtimeerror: expected a 'CUDA' device type for generator but found 'CPU'
How can I solve this problem？

您好，我使用最新版本的pytorch运行main()时，loss为nan，请问这是什么原因啊