I compare the origin repo with yours.

function ./pcn/pcn.py/stage1(img, imgPad, net, thres) line164.img_resized is changed by function preprocess_img in each loop. Is it correct?

I don't know whether the operator in the original repo img.convertTo(imgF, CV_32FC3); return imgF - mean; changes the source matrix img, but your code changes it. After the third loop, under most circumstances, np.max(img_resized) will be a negative integer.

Thank you.

root@ada39ed4587c:/app/pytorch-PCN# pip3 install .
WARNING: pip is being invoked by an old script wrapper. This will fail in a future version of pip.
Please see pypa/pip#5599 for advice on fixing the underlying issue.
To avoid this problem you can invoke Python with '-m pip' instead of running pip directly.
Processing /app/pytorch-PCN
ERROR: Command errored out with exit status 1:
command: /usr/bin/python3 -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-req-build-8u7yr5ph/setup.py'"'"'; file='"'"'/tmp/pip-req-build-8u7yr5ph/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' egg_info --egg-base /tmp/pip-req-build-8u7yr5ph/pip-egg-info
cwd: /tmp/pip-req-build-8u7yr5ph/
Complete output (7 lines):
Traceback (most recent call last):
File "", line 1, in
File "/tmp/pip-req-build-8u7yr5ph/setup.py", line 4, in
long_description = f.read()
File "/usr/lib/python3.6/encodings/ascii.py", line 26, in decode
return codecs.ascii_decode(input, self.errors)[0]
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe4 in position 19: ordinal not in range(128)
----------------------------------------
ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.

I get this error constantly. It seems like non ascii character in README.md

why process in pytorch 3-4s / 1 frame while in c++ 31frame / 1s???

hello sir
I don't know how to use this model. Can you give me a way to use it
An error occurred while calling the method and cannot be used. Can you help me

In utils.py class Window has only width parameter but not the height resulting in only square bounding boxes.

Hi, nice work. How about retraining on a new dataset? Because I find that the pre-trained PCN performs badly on low-resolution images.

How .pth files were obtained?

This is the loop for your stage1 which invokes first model repeatedly:

def stage1(img, imgPad, net, thres):
    row = (imgPad.shape[0] - img.shape[0]) // 2
    col = (imgPad.shape[1] - img.shape[1]) // 2
    winlist = []
    netSize = 24
    curScale = minFace_ / netSize
    img_resized = resize_img(img, curScale)
    while min(img_resized.shape[:2]) >= netSize:
        img_resized = preprocess_img(img_resized)   <---------------- !!!
        # net forward
        net_input = set_input(img_resized)
        with torch.no_grad():
            net.eval()
            cls_prob, rotate, bbox = net(net_input)

        w = netSize * curScale
        for i in range(cls_prob.shape[2]): # cls_prob[2]->height
            for j in range(cls_prob.shape[3]): # cls_prob[3]->width
                if cls_prob[0, 1, i, j].item() > thres:
                    sn = bbox[0, 0, i, j].item()
                    xn = bbox[0, 1, i, j].item()
                    yn = bbox[0, 2, i, j].item()
                    rx = int(j * curScale * stride_ - 0.5 * sn * w + sn * xn * w + 0.5 * w) + col
                    ry = int(i * curScale * stride_ - 0.5 * sn * w + sn * yn * w + 0.5 * w) + row
                    rw = int(w * sn)
                    if legal(rx, ry, imgPad) and legal(rx + rw - 1, ry + rw -1, imgPad):
                        if rotate[0, 1, i, j].item() > 0.5:
                            winlist.append(Window2(rx, ry, rw, rw, 0, curScale, cls_prob[0, 1, i, j].item()))
                        else:
                            winlist.append(Window2(rx, ry, rw, rw, 180, curScale, cls_prob[0, 1, i, j].item()))
        img_resized = resize_img(img_resized, scale_)  <---------------- !!!
        curScale = img.shape[0] / img_resized.shape[0]
    return winlist

I have marked repeated application of preprocess_img to img_resized. This function subtracts mean all the time so you can check that image mean shifts towards big negative numbers:

def preprocess_img(img, dim=None):
    if dim:
        img = cv2.resize(img, (dim, dim), interpolation=cv2.INTER_NEAREST)
    return img - np.array([104, 117, 123])

But you infer through the same network all the time, so this isn't possible that you have to feed different inputs.

I have moved preprocessing out of the loop, and the result is still ok. Is it a bug?

hi
what changes in code required to check several numbers of images output in one go.
not only one image

Hello. I tested it on the GPU, and it slowed down a lot. After debugging, I found that pcn.py→def stage1(),These two double "for" take a lot of time.
for i in range(cls_prob.shape[2]): # cls_prob[2]->height
for j in range(cls_prob.shape[3]): # cls_prob[3]->width
I've put net_input into the GPU, corresponding， CLS_ Prob, rotate and bbox are also loaded into the GPU.these two “for” need changed???
Or I use GPU in the wrong way. Can you help me

您好，我尝试了用gpu运行，速度变慢了。经过调试，我发现是pcn.py→def stage1()函数中的这个双for循环 占用了大量的时间，比在cpu上还要更久。
for i in range(cls_prob.shape[2]): # cls_prob[2]->height
for j in range(cls_prob.shape[3]): # cls_prob[3]->width
我把模型加载到了gpu，把net_input也加载到了gpu ，相应的CLS_ Prob, rotate and bbox都在gpu，这两句for有需要改的么。
还是说我使用gpu的方式不对呢？你可以帮帮我么

I want to figure out how to train PCN from wider face dataset.

I try the demo, found that it uses CPU only, how can I use GPU with the model?