SFD implement with pytorch

Hi, I have downloaded your ''face model'' which is provided in README, and successfully run it on my machine. But now I still have some questions because it seems works well on faces but not work well on heads. Here is my question: Is the "face model" trained with both head dataset "SCUT-HEAD" and face dataset "widerface", or just only trained with face dataset "widerface"?

I am finding it real hard to download checkpoint files from the website you uploaded in as I do not understand Chinese. Is there any other way you can share them? Much appreciated. Thanks.

I ran wider_test.py and got a folder —— s3fd_val in eval_tools. Then I ran wider_eval.m in MATLAB, but some errors were reported.

错误使用 legend (line 273) 未知的属性 'show'

出错 plot_pr (line 22) legend1 = legend(lendge_name,'show');

出错 wider_plot (line 33) plot_pr(propose, recall, name_list, seting_class, set_list{i},dateset_class);

出错 wider_eval (line 35) wider_plot(setting_name_list,dir_int,seting_class,dateset_class);

Hi, thank you very much for your open source code. However, I run the 'demo.py' and find the running on one frame using RTX 3080 is about 0.5 seconds. The original paper claimed this method is real-time.

Could you please give me some advice on this?

def nms(boxes, scores, overlap=0.5, top_k=200): keep = scores.new(scores.size(0)).zero_().long() if boxes.numel() == 0: return keep x1 = boxes[:, 0] y1 = boxes[:, 1] x2 = boxes[:, 2] y2 = boxes[:, 3] area = torch.mul(x2 - x1, y2 - y1) v, idx = scores.sort(0) # sort in ascending order # I = I[v >= 0.01] idx = idx[-top_k:] # indices of the top-k largest vals xx1 = boxes.new() yy1 = boxes.new() xx2 = boxes.new() yy2 = boxes.new() w = boxes.new() h = boxes.new()

# keep = torch.Tensor()
count = 0
while idx.numel() > 0:
    i = idx[-1]  # index of current largest val
    # keep.append(i)
    keep[count] = i
    count += 1
    if idx.size(0) == 1:
        break
    idx = idx[:-1]  # remove kept element from view
    # load bboxes of next highest vals
    torch.index_select(x1, 0, idx, out=xx1)
    torch.index_select(y1, 0, idx, out=yy1)
    torch.index_select(x2, 0, idx, out=xx2)
    torch.index_select(y2, 0, idx, out=yy2)
    # store element-wise max with next highest score
    xx1 = torch.clamp(xx1, min=x1[i])
    yy1 = torch.clamp(yy1, min=y1[i])
    xx2 = torch.clamp(xx2, max=x2[i])
    yy2 = torch.clamp(yy2, max=y2[i])
    w.resize_as_(xx2)
    h.resize_as_(yy2)
    w = xx2 - xx1
    h = yy2 - yy1
    # check sizes of xx1 and xx2.. after each iteration
    w = torch.clamp(w, min=0.0)
    h = torch.clamp(h, min=0.0)
    inter = w * h
    # IoU = i / (area(a) + area(b) - i)
    rem_areas = torch.index_select(area, 0, idx)  # load remaining areas)
    union = (rem_areas - inter) + area[i]
    IoU = inter / union  # store result in iou
    # keep only elements with an IoU <= overlap
    idx = idx[IoU.le(overlap)]
**_return keep, count_**

the two return are not same ,it may take error: layers/functions/detection.py", line 67, in forward ValueError: not enough values to unpack (expected 2, got 0)

Hi, Could you please provide the pretrained weight file for demo.py, so that I can test it on any image? Also the basenet vgg16_reducedfc.pth for training? Thanks.

Thanks for your code in the paper

• the cls loss is normalized by the number of positive and negative anchors
• the reg term is normalized by the number of positive anchors. so may be it should be changed

M = num_pos.data.sum().float() + num_neg.data.sum().float()
loss_c /= N

also in the paper L =\lambda * L_{cls} + L_{reg} in the code

loss_l, loss_c = criterion(out, targets)
loss = loss_l + loss_c

may be the two place should be modified

Whenever i am passing the images to the model,the memory consumption of RAM is increasing heavily. So is there any way to reduce it? or else this model itself need more configurations?

Thanks in advance.