search

lightaime / deep_gcns

Tensorflow Repo for "DeepGCNs: Can GCNs Go as Deep as CNNs?" ICCV2019 Oral https://www.deepgcns.org
MIT License
630 stars 89 forks source link

why the inpus_ph default to shape=(batch_size, num_vertices, 9)? #13

Closed zem007 closed 4 years ago

zem007 commented 4 years ago

Hi! As your write in _inputs_ph = tf.placeholder(tf.float32, shape=(batch_size, numvertices, 9)).

Why it is default to 9? where is the 9 from? I think the 3D point cloud should be 3.

If I wish to use the model to my own point cloud dataset with (x,y,z) coordinates => 3， should I change the original code to shape=(batch_size, num_vertices, 3)?

thanks!

lightaime commented 4 years ago

Hi @zem007 ,

The data we used has 9 dimensional features including 3D for xyz coordinates, 3D for RGB color and 3D for normalized xyz coordinates. Sure, you can chage the input shape to (batch_size, num_vertices, 3) to make the model fit for your data.

zem007 commented 4 years ago

Hi @zem007 ,

The data we used has 9 dimensional features including 3D for xyz coordinates, 3D for RGB color and 3D for normalized xyz coordinates. Sure, you can chage the input shape to (batch_size, num_vertices, 3) to make the model fit for your data.

thanks! It works well in my own dataset.