Training data acquisition method and system based on deep learning

Abstract

The invention discloses a training data acquisition method based on deep learning. The method comprises the steps of acquiring sample data, preprocessing the obtained sample data so as to obtain the preprocessed sample data, inputting the preprocessed sample data into a trained DCGAN model, and acquiring an output result as the final training data. The training data acquisition method can solve the technical problems that in an existing industrial internet identification analysis system, due to the fact that the number of label data is too large and the types are numerous, the label data used for training is difficult to collect, the labor cost is increased when a large amount of data is collected, time and labor are consumed, and overfitting occurs when a recognition model is trained due to the lack of diversity and high redundancy of the collected data.

Description

technical field [0001] The invention belongs to the field of computer application software, and more specifically relates to a training data acquisition method and system based on deep learning. Background technique [0002] In the industrial Internet system, the network is the foundation, and the logo is the foundation of the network and the "ID card" of the network. The industrial Internet logo analysis system is the key hub to realize the information exchange of all elements of the industry and all links; by giving each object With the help of the Industrial Internet logo analysis system, cross-regional, cross-industry, and cross-enterprise information query and sharing can be realized. [0003] At present, the identification solution system of the Industrial Internet assigns unique labels to items through barcodes, QR codes, and radio frequency identification tags. If you need to accurately identify the label, you need to use an efficient label recognition model, and th...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above defects or improvement needs of the prior art, the present invention provides a training data acquisition method and system based on deep learning. There are many technical problems that lead to difficulties in collecting label data for training, as well as technical problems that increase labor costs, time-consuming and labor-intensive collection of large amounts of data, and training recognition models due to lack of diversity and high redundancy in collected data. The technical problem of overfitting occurs when

Images