The present invention relates to a method and apparatus for providing a unified feature map using an ensemble of a plurality of outputs from a convolutional neural network. More specifically, the method comprising: (a) generating, by a CNN device, a plurality of transformed input images by receiving an input image, and applying a plurality of transform functions for transforming the input image; (b) obtaining, by the CNN device, each of the transformed feature maps corresponding to each of the transformed input images by applying a convolution operation to each of the transformed input images; (c) generating, by the CNN apparatus, an inverse transform function corresponding to each of the transform functions to each of the transformed feature maps to generate each inverse transform feature map corresponding to each of the transformed feature maps; and (d) obtaining, by the CNN apparatus, an integrated feature map by integrating at least some of the inverse transform feature maps.
展开▼
