Graph Convolutional Networks for Hyperspectral Image Classification
Citations
185 citations
173 citations
168 citations
Cites methods from "Graph Convolutional Networks for Hy..."
...That is, the graph is built by using a radial basis function (RBF) to measure the similarities among samples belonging to the same class [81],...
[...]
117 citations
96 citations
References
111,197 citations
"Graph Convolutional Networks for Hy..." refers methods in this paper
...Adam [43] is used to optimize the networks....
[...]
15,696 citations
"Graph Convolutional Networks for Hy..." refers background or methods in this paper
...Comparatively, graph convolutional networks (GCNs) [32] are a hot topic and emerging network architecture, which is able to effectively handle graph structure data by modeling relations between samples (or vertexes)....
[...]
...By limiting K = 1 and assigning the largest eigenvalue λmax of L̃ to 2 [32], (15) can be further simplified to...
[...]
...7) For the GCN, similar to [32], a graph convolutional hidden layer with 128 units is implemented in the GCN before feeding the features into the softmax layer, where the adjacency matrix à can be computed using KNN-based graph (K = 10 in our case)....
[...]
4,154 citations
2,373 citations
"Graph Convolutional Networks for Hy..." refers background in this paper
...However, the high spectral mixing between materials [5] and spectral variability and complex noise effects [6] bring difficulties in extracting discriminative information from such data....
[...]
2,059 citations
"Graph Convolutional Networks for Hy..." refers methods in this paper
...[27] adopted convolutional neural networks (CNNs) to extract spatial–spectral features more effectively from HS images, thereby yielding higher classification performance....
[...]