This paper proposes an autonomous focal location transition method that imitates saccadic eye movement for image recognition. Grid Cell Net (GCN) is an image recognition algorithm inspired by the human neocortex. GCN focuses on a part of the target image at the same time and sequentially changes the focal location on the target image. At each time step, GCN receives feature data from the focal location of the target image and the focal location transition. GCN recognizes the image from these time series data. However, the conventional GCN randomly transitions the focal locations to recognize the target image and lacks a mechanism to autonomously determine the focal location transition. The proposed method introduces an autonomic focal location transition to enhance image recognition by incorporating a candidate class selection and a next focal location prediction. The candidate class selection finds a promising recognition class at each transition step, while the next focal location prediction finds the next focal location that effectively reduces the recog-nition class candidates. Experimental results using the MNIST dataset show that the proposed method not only autonomously transitions the focal location but also achieves higher recognition accuracy compared to the conventional GCN.