Combining the global features from PointNet with the local features from CNN

Hi @miguelriemoliveira and @pmdjdias,

today I had an idea that we could explore: build a neural network that receives both the point cloud and gray image. A sketch of the architecture is the following:

Using the point cloud we can extract the global features and using the gray image we can extract local features. By merging both information, maybe we can achieve something...

Of course, before entering this we first need to explore both modalities individually, but this could be a good place to innovate.

DanielCoelho112 / synfeal

Combining the global features from PointNet with the local features from CNN #20