[SOLVED] How to Implement DSFD on Windows, Mac, and Linux

[david-littlefield]

Here's an article I wrote that shows how to get the demo working on Windows, Mac, and Linux

[berkanttubi]

Thank you so much david, however I faced anouther error too. Can you have any solution about it?

test_oneimage()
  File "/home/berkanttugberk/GithubProjects/FaceDetection-DSFD/demo.py", line 190, in test_oneimage
    det0 = infer(net , img , transform , thresh , cuda , shrink)
  File "/home/berkanttugberk/GithubProjects/FaceDetection-DSFD/demo.py", line 71, in infer
    y = net(x)      # forward pass
  File "/home/berkanttugberk/miniconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 727, in _call
_impl
    result = self.forward(*input, **kwargs)
TypeError: forward() missing 1 required positional argument: 'x'

[david-littlefield]

Hi-there @berkanttubi. Is that the entire error message?

[berkanttubi]

Hi-there @berkanttubi. Is that the entire error message?

Here is the entire message;

[david-littlefield]

@berkanttubi You will need to make the changes that are covered in the article to prevent these errors.

[berkanttubi]

@david-littlefield I did :)

[david-littlefield]

The first step in the "Fix the Errors" section should solve this problem.

Can you send a screenshot of your code from the stated files at the stated lines of code?

[berkanttubi]

@david-littlefield I think the problem was about the torch version. When I downgraded it, the code worked. Thank you so much for your article David. I couldn't do it without your instructions. But I have another question, It seems the model is very slow. I tried to work with 30 fps 30 sec video, and it took 280 sec to detect it. Isn't it a bit slow? What should fps be, any idea?

[david-littlefield]

@berkanttubi Good. Glad you were able to get it working! I've only used DSFD with single images, but I vaguely remember the paper stating something about the model not being fast enough to handle live video. What kind of GPU are you using?

[afnan-shazab]

im getting the following error loading pretrained resnet model Finished loading model! Error bbox number to filter : 0, bbox number: 0 Finished loading data loading pretrained resnet model Finished loading model! Traceback (most recent call last): File "demo.py", line 232, in test_oneimage() File "demo.py", line 201, in test_oneimage det0 = infer(net , img , transform , thresh , cuda , shrink) File "demo.py", line 80, in infer y = net(x) # forward pass File "/home/shazab/miniconda3/envs/dsfd/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl return forward_call(*input, **kwargs) File "/home/shazab/FaceDetection-DSFD/face_ssd.py", line 348, in forward output = self.detect.forward( File "/home/shazab/FaceDetection-DSFD/layers/functions/detection.py", line 75, in forward ids, count = nms(boxes, scores, self.nms_thresh, self.top_k) File "/home/shazab/FaceDetection-DSFD/layers/box_utils.py", line 369, in nms union = (rem_areas - inter) + area[i] RuntimeError: The size of tensor a (4999) must match the size of tensor b (0) at non-singleton dimension 0

i dont know what to do