We propose a multi-zone blocking model to obtain the chunking area of the image, and then locate the local detail features through the spatial transformer network.
We propose a multi region block model to obtain the block region of the image, and then locate the local details through the spatial transformer network.<br>