Can deep_gcn take point cloud repressed by 3d coordinate only?

I went through your paper and code and found each point has more than 3 features. However, could I take each point with xyz only and NOT including rest features like surface normal and color? Appreciate if you can provide more instruction how could I adjust the number of input channel. Thank you

Sure you can. You can simply set the input channels as 3 by --in_channels 3. See

deep_gcns_torch/examples/sem_seg_dense/opt.py

Line 24 in da2bdd6

    
           parser.add_argument('--in_channels', default=9, type=int, help='the channel size of input point cloud ')

and

deep_gcns_torch/examples/sem_seg_dense/architecture.py

Line 21 in da2bdd6

self.head = GraphConv2d(opt.in_channels, channels, conv, act, norm, bias)

.

Sure you can. You can simply set the input channels as 3 by --in_channels 3. See

deep_gcns_torch/examples/sem_seg_dense/opt.py

Line 24 in da2bdd6

parser.add_argument('--in_channels', default=9, type=int, help='the channel size of input point cloud ')

and

deep_gcns_torch/examples/sem_seg_dense/architecture.py

Line 21 in da2bdd6

self.head = GraphConv2d(opt.in_channels, channels, conv, act, norm, bias)

.

Thank you for the quick response. Please stay safe recently :)

Thanks. Stay safe!