vidursatija / blazepalm Goto Github PK

It's a great share. I have a question about dataset. How can we get the dataset hand landmarks. Could you share it if it is a public dataset?

Hi,

When I run the run_detector.py, I got the following errors:

Traceback (most recent call last):
File "C:\Users\nieshijie\Downloads\BlazePalm-master\BlazePalm-master\ML\run_detector.py", line 31, in
cropped_hand = frame[y:endy, x:endx]
TypeError: only integer tensors of a single element can be converted to an index

I modified line 25~30 and convert type from float to int. It seems that the detected bound box is wrong. Could you please check about that? Thanks.

Hello, I want to know the training data set of Hand model, can you provide it to me?

thank you for your great works!
I have a question about your work.

1. why the size of tflite is different from media-pipe's one?

Hello, I get this error when I ran run_detector.py,I want to know how to solve it

/home//zhouyan1/anaconda3/envs/gesture_recognition/bin/python /home/SENSETIME/zhouyan1/PycharmProjects/BlazePalm-master/ML/run_detector.py
Traceback (most recent call last):
File "/home//zhouyan1/PycharmProjects/BlazePalm-master/ML/run_detector.py", line 39, in
_, _, landmarks = hl(torch.from_numpy(cropped_hand).permute((2, 0, 1)).unsqueeze(0))
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home//zhouyan1/PycharmProjects/BlazePalm-master/ML/handlandmarks.py", line 115, in forward
b1 = self.backbone1(x) # 64x64
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/conv.py", line 446, in forward
return self._conv_forward(input, self.weight, self.bias)
File "/home//zhouyan1/anaconda3/envs/gesture_recognition/lib/python3.9/site-packages/torch/nn/modules/conv.py", line 442, in _conv_forward
return F.conv2d(input, weight, bias, self.stride,
RuntimeError: expected scalar type Byte but found Float

Hi,

• Running run_detector.py program on Ubuntu 22.04.2 LTS.
• Pytorch version is 2.0.1+cpu

Initially, I received this issue. I was able to fix that issue by following the solution mentioned in the issue.

Now, I am getting this error:

aravind@aravind-Latitude-E6420:ML$ python3 run_detector.py 
[W NNPACK.cpp:64] Could not initialize NNPACK! Reason: Unsupported hardware.
Warning: Ignoring XDG_SESSION_TYPE=wayland on Gnome. Use QT_QPA_PLATFORM=wayland to run on Wayland anyway.
Traceback (most recent call last):
  File "/home/aravind/work/pp/machine_learning/BlazePalm/ML/run_detector.py", line 33, in <module>
    cropped_hand = np.pad(cropped_hand,
  File "<__array_function__ internals>", line 200, in pad
  File "/home/aravind/.local/lib/python3.10/site-packages/numpy/lib/arraypad.py", line 744, in pad
    pad_width = _as_pairs(pad_width, array.ndim, as_index=True)
  File "/home/aravind/.local/lib/python3.10/site-packages/numpy/lib/arraypad.py", line 518, in _as_pairs
    return np.broadcast_to(x, (ndim, 2)).tolist()
  File "<__array_function__ internals>", line 200, in broadcast_to
  File "/home/aravind/.local/lib/python3.10/site-packages/numpy/lib/stride_tricks.py", line 413, in broadcast_to
    return _broadcast_to(array, shape, subok=subok, readonly=True)
  File "/home/aravind/.local/lib/python3.10/site-packages/numpy/lib/stride_tricks.py", line 349, in _broadcast_to
    it = np.nditer(
ValueError: operands could not be broadcast together with remapped shapes [original->remapped]: (2,2)  and requested shape (3,2)

Please help.

Thank you,
Aravind.

It's a great share. I have a question about dataset. How can we get the dataset hand landmarks. Could you share it if it is a public dataset?

Hello, I tried to run BlazePalm on iPhone 7 and found 2 problems:

1. The detection accuracy of BlackePalm for hand key points is worse than that of MediaPipe Hands, which sometimes leads to false recognition.
2. MediaPipe Hands on iPhone 7 can run at a performance close to 30FPS, but BlackePalm seems to have less than 15FPS.
   Can the above problems be optimized? Is there a solution? be deeply grateful!
   
   

Could the input size be changed?