A GIF format picture retrieval method based on
machine learning comprises the following steps that S1, obtaining a GIF picture in a network, and preprocessing the GIF picture; s2,
cutting the preprocessed GIF picture, extracting a
key frame in the GIF picture, calculating a Hash value of the
key frame, and storing the generated Hash value information and the GIF picture in a basic
database; s3, constructing a
convolutional neural network model, training the
convolutional neural network model by using data in the basic
database, and obtaining a kernel function for prediction; s4, splitting the to-be-searched target GIF picture, extracting a
key frame, and substituting the extracted key frame into a kernel function for calculation to obtain an
optimal matching type; and S5, calculating a hash value of a key frame of the to-be-searched target GIF picture, searching similar pictures in the predicted types by using the hash value information, and if not, creating a new category in the current category. According to the picture searching function, the GIF picture searching function is added, the searching accuracy is improved, and finally the
utilization rate of picture searching is increased.