Neural network based recognition apparatus and method of training neural network
A neural network recognition method includes obtaining a first neural network that includes layers and a second neural network that includes a layer connected to the first neural network, actuating a processor to compute a first feature map from input data based on a layer of the first neural network, compute a second feature map from the input data based on the layer connected to the first neural network in the second neural network, and generate a recognition result based on the first neural network from an intermediate feature map computed by applying an element-wise operation to the first feature map and the second feature map.
1. A processor implemented neural network recognition method, comprising:
obtaining a first neural network comprising layers and a second neural network comprising a layer connected to the first neural network;
determining a first feature map from input data based on a layer of the first neural network;
determining a second feature map from the input data based on the layer connected to the first neural network in the second neural network; and
generating a recognition result based on the first neural network from an intermediate feature map determined by applying an element-wise operation to the first feature map and the second feature map.
2. The method of claim 1 , wherein the determining of the first feature map comprises determining the first feature map corresponding to the input data based on a previous layer of a target layer included in the first neural network.
3. The method of claim 2 , wherein the generating of the recognition result includes generating the recognition result from the intermediate feature map based on a next layer of the target layer included in the first neural network.
4. The method of claim 1 , wherein the determining of the second feature map comprises determining the second feature map corresponding to the input data based on a layer connected to a target layer included in the first neural network, among a plurality of layers included in the second neural network, and providing the second feature map to the first neural network.
5. The method of claim 1 , further comprising:
preprocessing the second feature map and providing the preprocessed second feature map to the first neural network.
6. The method of claim 1 , further comprising:
generating a recognition result from the input data based on the second neural network.
7. The method of claim 1 , wherein a total number of nodes included in a layer of the first neural network is equal to a total number of nodes included in the layer connected to the first neural network.
8. The method of claim 1 , further comprising:
determining a third feature map corresponding to at least one of plural layers in the first neural network, and providing the third feature map to a third neural network to generate a recognition result with respect to the third neural network.
9. The method of claim 1 , further comprising:
determining a feature map of a target layer included in the first neural network based on the target layer from a feature map of a previous layer included in the first neural network in response to the target layer being connected to the previous layer.
10. The method of claim 1 , wherein the generating of the recognition result comprises:
performing the applying of the element-wise operation to the first feature map and the second feature map by applying the element-wise operation to an individual element of the first feature map and an element corresponding to the individual element in the second feature map; and
generating the intermediate feature map based on results of the performed applying.
11. A non-transitory computer-readable storage medium storing program instructions that, when executed by a processor, cause the processor to perform the method of claim 1 .