Free Access

Fig. 2


Example of a CNN architecture: The input image undergoes a series of convolution layers into a series of feature maps. The first convolution transforms the 101 × 101 pixel image into four 101 × 101 pixel feature maps. To lower computation cost, max-pooling layers are used in between convolutions. They reduce the dimensionality of the image, dividing the size of the image by two. A fully connected layer then combines all feature maps for the classification.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.