| CPC G06N 3/082 (2013.01) | 9 Claims |
|
1. A method for anomaly detection comprising:
preparing, by a learning unit, training input data by selecting data in a normal state;
entering, by the learning unit, the training input data into a detection network;
generating, by the detection network, an attention map by performing a plurality of operations in which a plurality of layer weights are applied to the training input data, followed by deriving training output data imitating the training input data;
calculating, by the learning unit, a restoration loss indicating a difference between the training input data and the training output data;
performing, by the learning unit, optimization for updating a weight of the detection network through a backpropagation algorithm to reduce the restoration loss;
entering, by the detection unit, input data into a detection network;
generating, by the detection network, an attention map and output data through a plurality of operations in which a plurality of layer weights are applied to the input data;
generating, by the detection unit, an attention map by overlapping the attention map with the input data when an attention region having an attention value greater than or equal to a predetermined threshold exists in the attention map;
detecting, by the detection unit, whether the input data is normal or abnormal according to the output data; and
outputting, by the detection unit, the detection map and an anomaly detection result indicating whether the input data is normal or abnormal.
|