A detection system and method with data recognition function

Abstract

The invention proposes a detection system and method with data identification function. The system includes: a data sampling subsystem, which samples data from the input data stream to obtain multiple sampled data sets; a target detection network training subsystem, which trains multiple target detection networks based on multiple sampled data sets; a target detection network training subsystem Output multiple groups of configuration parameter sets of multiple deep learning neural network models; the method includes: input multiple groups of configuration parameter sets into the target detection network evaluation model for prediction, and obtain a network prediction performance parameter set; determine that the network prediction performance parameter set satisfies For the first network prediction performance parameter set of the first performance condition, the first network prediction performance parameter set is used as an adjustment set, and the parameters of the target detection network evaluation model are adjusted to obtain an adjusted target detection network evaluation model. The invention also proposes a computer-readable storage medium implementing the method.

Description

technical field [0001] The invention belongs to the field of data identification and detection, and in particular relates to a detection system and method with data identification function, and a computer-readable storage medium. Background technique [0002] Using deep learning to automatically learn features has gradually replaced manual construction of features and statistical methods. But one of the key problems is that a large amount of data is required, otherwise it will be overfitting due to too many parameters. But this cost is very high, so obtaining limited data from massive big data streams to train efficient deep learning models has become a problem that needs to be solved. An important milestone is transfer learning—inspired by humans, instead of learning from scratch from a large amount of data, it uses a small number of examples to solve a problem. [0003] Due to the continuous popularization of the network and related applications, network data is graduall...

AI Technical Summary

Problems solved by technology

But this cost is very high, so obtaining limited data from massive big data streams to train efficient deep learning models has become a problem that needs to be solved

Images