Test set compression method, system and device and storage medium

The invention discloses a test set compression method, system and device and a storage medium, and belongs to the technical field of deep learning and data compression. The method comprises the following steps: acquiring a test set, and extracting samples from the test set by adopting an HAST model; calculating the complexity d of the extracted sample x by adopting GHM; dividing the samples into complex samples and simple samples according to the complexity d to obtain a complex sample set and a simple sample set; and respectively extracting samples from the complex sample set and the simple sample set according to a preset proportion to form a compressed test set. According to the method, a complete test set can be effectively compressed through a compression algorithm of complex sample extraction designed by using the HAST model, the most representative and challenging sample is reserved, and compression ratio quantification at different mixing precisions can be quickly and accurately completed to optimize