A multi-detector SPECT system equipped with multi-pinhole collimators has great advantages compared with a conventional one owing to a data acquisition without rotating a gamma camera. It reduces the data acquisition time and obtains dynamic information of tracers. However, it has a problem in the image reconstruction; reconstructed images are highly degraded when the size of a pinhole is large. In this study, we propose a deeplearning based image-reconstruction method for a multi-pinhole static SPECT system.