The invention discloses a collaborative attention VQA method and device based on grids, and belongs to the field of visual questions and answers. The collaborative attention VQA method comprises the following steps: acquiring an image from a data set and carrying out mesh generation; inputting the image after grid division into an RCNN, and obtaining a feature map after convolution, pooling and feature fusion; obtaining a problem from the data set and mapping the problem into a vector space to obtain a word vector; calculating a correlation matrix according to the feature map and the word vector, and calculating attention distribution of the feature map and attention distribution of the word vector; inputting GRU to calculate to obtain a new word vector and a corresponding code; and integrating the new word vector and the corresponding code by adopting the MLP to obtain an answer corresponding to the question. The collaborative attention VQA device comprises a division module, an RCNNmodule, a mapping module, a calculation module, a GRU module and an MLP module. The collaborative attention VQA method and device can realize mutual attention of the image and the problem text, can improve the prediction accuracy, and can improve the performance of the model.