An image detection-based fine-grained classification model processing method, belonging to the field of artificial intelligence, and comprising: receiving a keyword, and constructing an image data set by means of a search engine; randomly grouping the image data set into several training sets; inputting the several training sets into a fine-grained classification initial model, and obtaining attention weight vectors of images in the several training sets; pooling the attention weight vectors, and respectively generating training instances corresponding to the several training sets; inputting the training instances into a classifier of the fine-grained classification initial model, so as to calculate a model loss; adjusting model parameters according to the model loss, and obtaining a fine-grained classification model. Also provided are an image detection-based fine-grained classification model processing apparatus, a computer device, and a storage medium. In addition, the present invention relates to blockchain technology, and trained model parameters may be stored in a blockchain. The present invention rapidly and accurately implements fine-grained image classification processing.