Abstract: In a convolutional neural network (CNN) using an encoder-decoder structure for image segmentation, a multi-scale context aggregation module receives an encoded final-stage feature map from the encoder, and sequentially aggregates multi-scale contexts of this feature map from a global scale to a local scale to strengthen semantic relationships of contexts of different scales to improve segmentation accuracy. The multi-scale contexts are obtained by computing atrous convolution on the feature map for different dilation rates. To reduce computation, a channel-wise feature selection (CFS) module is used in the decoder to merge two input feature maps. Each feature map is processed by a global pooling layer followed by a fully connected layer or a 1×1 convolutional layer to select channels of high activation. By subsequent channel-wise multiplication and elementwise summation, only channels with high activation in both feature maps are preserved and enhanced in the merged feature map.

Abstract: Polarization-independent focusing is advantageously achieved by a multifocal system having a polarization beam splitter (PBS) to split an unpolarized light beam into two orthogonally linearly-polarized (LP) light beams. The two LP light beams are reflected by mirrors to travel in opposite directions and enter into a variable-focusing module at two ends thereof, respectively. The module includes waveplates to convert the LP light beams into two circularly-polarized (CP) light beams at both ends of an optical assembly. The optical assembly is formed with a stack of birefringent optical elements including at least one geometric phase lens and one polarization selector that may be electrically modulated to select the optical power in focusing the two CP light beams. Followed by the waveplates converting two focused CP light beams to two focused LP light beams and upon mirror reflection, the beams are finally recombined by the PBS to form one focused light beam.