Erratum to “Multiscale Fully Convolutional Network for Foreground Object Detection in Infrared Videos” [Apr 18 617-621]
Dongdong Zeng, Ming Lei Zhu · IEEE Geoscience and Remote Sensing Letters · 2018
The architecture of the proposed multiscale fully convolutional network (MFCN) in our paper[1]is mainly derived from a salient object method[2]and a semantic segmentation method[3]. We missed these two reference papers in the original paper. The MFCN is an encoder-decoder architecture; the output has the same size with the input. For the encoder part, we use the pretrained VGG16 network, and for the decoder part, we upsample the features with deconvolution operations. The contrast layer for feature extraction is derived from the salient detection method[2].