Board damage classification system
A board damage classification system includes a Convolutional Neural Network (CNN) sub-engine and a Graph Convolutional Network (GCN) sub-engine that were trained based on digital images of structures that have experienced natural disasters. The CNN sub-engine receives a board digital image of a board, analyzes the board digital image to identify board features, and determines a board feature damage classification for the board features. The CGN sub-engine receives a board feature graph that was generated using the board digital image and that includes nodes that correspond to the board features in the board digital image, and defines relationships between the nodes included in the board feature graph. The board feature damage classification determined by the CNN sub-engine and the relationships defined by the GCN sub-engine are then used to generate a board damage classification that includes a damage probability for board features in the board digital image.
1. A board damage classification system, comprising:
a processing system; and
a memory system that is coupled to the processing system and that includes instructions that, when executed by the processing system, cause the processing system to provide a board damage classification engine that includes:
a Convolutional Neural Network (CNN) sub-engine that was trained to identify board damage using training digital images of structures that have experienced natural disasters, wherein the CNN sub-engine is configured to:
receive a board digital image of a board;
analyze the board digital image to identify board features; and
determine a board feature damage classification for at least one of the board features identified in the board digital image; and
a Graph Convolutional Network (GCN) sub-engine that was trained to identify board feature damage relationships using training graphs generated from the training digital images of structures that have experienced natural disasters, wherein the CGN sub-engine is configured to:
receive a board feature graph that was generated using the board digital image and that includes a plurality of nodes that correspond to the board features in the board digital image; and
define at least one relationship between the plurality of nodes included in the board feature graph,
wherein the board damage classification engine is configured to:
utilize the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate a board damage classification that includes a damage probability for at least one of the board features in the board digital image; and
provide the board damage classification for display.
2. The system of claim 1 , wherein the training digital images of structures that have experienced natural disasters include structure-before-natural-disaster digital images, structure-after-natural-disaster digital images, and structure-masked-labeled digital images.
3. The system of claim 1 , wherein the board damage classification engine includes:
a graph generation sub-engine that is configured to:
receive the digital image of the board;
generate the board feature graph using the digital image of the board; and
provide the board feature graph to the GCN sub-engine.
4. The system of claim 1 , wherein the board feature damage classification for the at least one of the board features includes one of a not-damaged board feature classification, a minor-damage board feature classification, a major-damage board feature classification, or a destroyed board feature classification.
5. The system of claim 1 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
concatenating the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine into a plurality of hidden layers; and
generating a fully connected layer from the plurality of hidden layers.
6. The system of claim 5 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
generating, via a softmax layer, the damage probability for at least one of the board features in the board digital image based on the fully connected layer.
7. An Information Handling System (IHS), comprising:
a processing system; and
a memory system that is coupled to the processing system and that includes instructions that, when executed by the processing system, cause the processing system to provide a board damage classification engine that includes:
a Convolutional Neural Network (CNN) sub-engine that is configured to:
receive a board digital image of a board;
analyze the board digital image to identify board features; and
determine a board feature damage classification for at least one of the board features identified in the board digital image; and
a Graph Convolutional Network (GCN) sub-engine that is configured to:
receive a board feature graph that was generated using the board digital image and that includes a plurality of nodes that correspond to the board features in the board digital image; and
define at least one relationship between the plurality of nodes included in the board feature graph,
wherein the board damage classification engine is configured to:
utilize the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate a board damage classification that includes a damage probability for at least one of the board features in the board digital image; and
provide the board damage classification for display.
8. The IHS of claim 7 , wherein the CNN sub-engine was trained to identify board damage using training digital images of structures that have experienced natural disasters, and wherein the GCN sub-engine was trained to identify board feature damage relationships using training graphs generated from the training digital images of structures that have experienced natural disasters.
9. The IHS of claim 8 , wherein the training digital images of structures that have experienced natural disasters include structure-before-natural-disaster digital images, structure-after-natural-disaster digital images, and structure-masked-labeled digital images.
10. The IHS of claim 7 , wherein the board damage classification engine includes:
a graph generation sub-engine that is configured to:
receive the digital image of the board;
generate the board feature graph using the digital image of the board; and
provide the board feature graph to the GCN sub-engine.
11. The IHS of claim 7 , wherein the board feature damage classification for the at least one of the board features includes one of a not-damaged board feature classification, a minor-damage board feature classification, a major-damage board feature classification, or a destroyed board feature classification.
12. The IHS of claim 7 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
concatenating the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine into a plurality of hidden layers; and
generating a fully connected layer from the plurality of hidden layers.
13. The IHS of claim 12 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
generating, via a softmax layer, the damage probability for at least one of the board features in the board digital image based on the fully connected layer.
14. A method for classifying damage to a board, comprising:
receiving, by a Convolutional Neural Network (CNN) sub-engine included in a board damage classification engine, a board digital image of a board;
analyzing, by the CNN sub-engine, the board digital image to identify board features; and
determining, by the CNN sub-engine, a board feature damage classification for at least one of the board features identified in the board digital image;
receiving, by a Graph Convolutional Network (GCN) sub-engine included in the board damage classification engine, a board feature graph that was generated using the board digital image and that includes a plurality of nodes that correspond to the board features in the board digital image;
defining, by the GCN sub-engine, at least one relationship between the plurality of nodes included in the board feature graph,
utilizing, by the board damage classification engine, the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate a board damage classification that includes a damage probability for at least one of the board features in the board digital image; and
providing, by the board damage classification engine, the board damage classification for display.
15. The method of claim 14 , further comprising:
training, by the CNN sub-engine, to identify board damage using training digital images of structures that have experienced natural disasters; and
training, by the GCN sub-engine, to identify board feature damage relationships using training graphs generated from the training digital images of structures that have experienced natural disasters.
16. The method of claim 15 , wherein the training digital images of structures that have experienced natural disasters include structure-before-natural-disaster digital images, structure-after-natural-disaster digital images, and structure-masked-labeled digital images.
17. The method of claim 14 , further comprising:
receiving, by a graph generation sub-engine included in the board damage classification engine, the digital image of the board;
generating, by the graph generation sub-engine, the board feature graph using the digital image of the board; and
providing, by the graph generation sub-engine, the board feature graph to the GCN sub-engine.
18. The method of claim 14 , wherein the damage classification for the at least one of the board features includes one of a not-damaged board feature classification, a minor-damage board feature classification, a major-damage board feature classification, or a destroyed board feature classification.
19. The method of claim 14 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
concatenating the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine into a plurality of hidden layers; and
generating a fully connected layer from the plurality of hidden layers.
20. The method of claim 14 , wherein the utilization of the board feature damage classification determined by the CNN sub-engine and the at least one relationship defined by the GCN sub-engine to generate the board damage classification includes:
generating, via a softmax layer, the damage probability for at least one of the board features in the board digital image based on the fully connected layer.