bug when running python kinect2grasp_python2.py

[TheJasal]

Traceback (most recent call last):
File "kinect2grasp_python2.py", line 51, in
from main_test import test_network, model, args
File "/main_test.py", line 50, in
model = torch.load(args.load_model, map_location='cpu')
File "/usr/local/lib/python2.7/dist-packages/torch/serialization.py", line 358, in load
return _load(f, map_location, pickle_module)
File "/usr/local/lib/python2.7/dist-packages/torch/serialization.py", line 542, in _load
result = unpickler.load()
AttributeError: 'module' object has no attribute '_rebuild_parameter'

how to fix this problem?thanks

[ruanjian0821]

This may be helpful.
NVIDIA/tacotron2#182

[TheJasal]

using torch 1.1.0 and referencing the answer by @ruanjian0821 solved the problem.
new issue:
Traceback (most recent call last):
File "kinect2grasp_python2.py", line 537, in
if_good_grasp, grasp_score_tmp = test_network(model.eval(), points_modify)
File "main_test.py", line 71, in test_network
output, _ = model_(local_pc) # N*C
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "pointnet.py", line 186, in forward
x, trans = self.feat(x)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "pointnet.py", line 137, in forward
trans = self.stn(x)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/home/robot/code/grasp-pointnet/PointNetGPD/model/pointnet.py", line 30, in forward
x = F.relu(self.bn1(self.conv1(x)))
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/conv.py", line 196, in forward
self.padding, self.dilation, self.groups)
RuntimeError: Calculated padded input size per channel: (0). Kernel size: (1). Kernel size can't be greater than actual input size.

[ruanjian0821]

Checking the "local_pc" shape.
It's looks like there is a problem with the input.

[TheJasal]

problem solved.
we use pointgpd_chann3_local.model as trained model
in main_test.py
change parser.add_argument('--model_type', type=str) to parser.add_argument('--model_type', type=str,default='100').

[ruanjian0821]

good news！

[TheJasal]

[TheJasal]

The calculated marker array is really high as the picture shows.
I don't know why.
I am sure the frames 'world','table_top' and 'kinect2_ir_optical_frame' are set correctly.
I checked the value of 'position' in real_good_grasp,and its z value is always 2 to 4,much higher than the ground.

[ruanjian0821]

Find in dex-net/apps/kinect2grasp_python2.py, some code may be helpful.
as the kinect2 is not well colibrated, here is a work around
points_[:, 0] = points_[:, 0] + 0.025
points_[:, 2] = points_[:, 2] #+ 0.018

[TheJasal]

The table_top_points and table_top topic don't know how to defined!

[trungpham2606]

@ruanjian0821
@TheJasal
Can you guys send me the weight: pointgpd_chann3_local.model ?
The new weight that the author provide seems not working.

[ACaseOfOrange]

The calculated marker array is really high as the picture shows.
I don't know why.
I am sure the frames 'world','table_top' and 'kinect2_ir_optical_frame' are set correctly.
I checked the value of 'position' in real_good_grasp,and its z value is always 2 to 4,much higher than the ground.

how did you set 'table_top' and 'kinect2_ir_optical_frame'? Thanks!

[Twilight89]

@TheJasal Hi~ Do you find the solution? I'm trouble in the same problem...