Thank you for your brilliant work.
Recently, I've tried to reproduce pointcnn on a pointcloud classification work. I did what the README instruction said about the dataset preparation.
But, when I run the train_val_dataset.sh, the training log always report the error as below:
File "../train_val_cls.py", line 344, in
main()
File "../train_val_cls.py", line 174, in main
net = model.Net(points=points_augmented, features=features_augmented, is_training=is_training, setting=setting)
File "/home/amax/PointCloud/PointCNN/pointcnn_cls.py", line 12, in init
PointCNN.init(self, points, features, is_training, setting)
File "/home/amax/PointCloud/PointCNN/pointcnn.py", line 116, in init
depth_multiplier, sorting_method, with_global)
File "/home/amax/PointCloud/PointCNN/pointcnn.py", line 23, in xconv
nn_fts_from_pts_0 = pf.dense(nn_pts_local, C_pts_fts, tag + 'nn_fts_from_pts_0', is_training)
File "/home/amax/PointCloud/PointCNN/pointfly.py", line 346, in dense
reuse=reuse, name=name, use_bias=not with_bn)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/util/deprecation.py", line 324, in new_func
return func(*args, **kwargs)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/layers/core.py", line 188, in dense
return layer.apply(inputs)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/keras/engine/base_layer.py", line 1479, in apply
return self.call(inputs, *args, **kwargs)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/layers/base.py", line 537, in call
outputs = super(Layer, self).call(inputs, *args, **kwargs)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/keras/engine/base_layer.py", line 586, in call
self.name)
File "/home/amax/venv/lib/python3.5/site-packages/tensorflow/python/keras/engine/input_spec.py", line 111, in assert_input_compatibility
layer_name + ' is incompatible with the layer: '
ValueError: Input 0 of layer xconv_1_nn_fts_from_pts_0 is incompatible with the layer: its rank is undefined, but the layer requires a defined rank.
It seems the input can't be loaded correctly and there are none issue said anything about this question. I would like to get some advice from you guys, thanks.
Hi @Viotho
It seems like you define a new pointcnn network structure and we can't reproduce your result. It's better to upload your config information to solve this issue.
Thanks.
Hi @burui11087 , I got the same exact error. I didn't change anything, I just simply clone the repository and follow the instruction without changing anything.
I'm using TF-1.14 on RTX 2070 with CUDA 10.0 and CuDNN 7.5 on Ubuntu 18.04. Please help, thanks.
Edit: I found the solution, downgrade to TF-1.13.2
Downgrading to TF-1.13.2 worked perfectly! Thanks @ronykalfarisi!
yangyanli
