A method and a system for image classification

Abstract

Disclosed is an apparatus for image classification. The apparatus comprises a converter and a forward propagator. The converter is configured to retrieve a convolutional neural network with a plurality of convolutional layers and a plurality of pooling layers connected to the convolutional layers. The forward propagator is configured to feed an image into the convolutional neural network to predict classes of all pixels in the image. The convert further comprises first and second converting units. The first converting unit is configured to insert all-zero rows and columns to the convolutional kernel of the convolutional layers such that every two neighboring entries are separated from each other. The second converting unit is configured to insert unmasked rows and columns to the pooling kernel of the pooling layers such that every two neighboring entries are separated from each other. The apparatus also comprises a backward propagator to update the convolutional kernels in the converted convolutional neural network. The present application also discloses a method for image classification.

Description

technical field [0001] The present application relates to methods and systems for image classification. Background technique [0002] The goal of pixel-wise classification is to classify all pixels in an image into different classes. Pixel-wise classification tasks include image segmentation and object detection, which require inputting image patches into a classifier and outputting a class label for the central pixel. [0003] Convolutional neural networks (CNNs), which are trainable multi-level feed-forward neural networks, have been extensively studied to extract good multi-level feature representations for image classification tasks. The input and output of each layer are called feature maps. CNN generally includes convolutional layers, pooling layers and nonlinear layers. The convolutional layer performs a convolution operation on the input feature map with a 3D filter bank to generate an output feature map. Each filter extracts the same type of local features at al...

AI Technical Summary

Problems solved by technology

It is extremely inefficient to directly apply it to pixel-by-pixel classification in a block-by-block scanning manner, because the surrounding blocks of pixels have large overlaps, resulting in many redundant calculations

Images