Data enhancement learning and training method, electronic equipment and readable storage medium

Abstract

The embodiment of the invention provides a data enhancement learning and training method, electronic equipment and a readable storage medium. The method comprises the following steps: mixing a staticimage into each frame of a sample video according to a scale factor; according to the embodiment of the invention, a data enhancement method TCA is utilized to guide the learning target of the whole neural network; the TCA can be simply integrated in any neural network, specifically, a static image is mixed into each frame of a sample video according to a scale factor, and the similarity of time clues under different space contexts can be reserved by selecting a proper scale factor. In addition, the TCA can be realized through simple matrix operation, the calculation overhead is very low, themethod provided by the embodiment of the invention achieves the optimal effect on three data sets, the effectiveness of the data enhancement method is verified, the TCA avoids the requirement on a real label, and the method can be expanded to self-supervised and semi-supervised learning.

Description

technical field [0001] The invention relates to the technical field of video processing, in particular to a data reinforcement learning, a training method, an electronic device, and a readable storage medium. Background technique [0002] Data Augmentation is a very common technique in deep learning. In image classification, the input image is usually elastically deformed or noise is added, which can greatly change the pixel content of the image without changing the label. Based on this, many enhancement techniques for rotation are proposed, such as flipping and color dithering. Data augmentation can improve the diversity of samples and greatly improve the robustness of the model. [0003] The existing MixUp is a practical image classification data enhancement method, and its effectiveness has been verified in the image-based field. For the samples in the data set, in the training process, all samples are first divided into different batches and randomly Sample one of the...

AI Technical Summary

Problems solved by technology

However, the significant disadvantage of MixUp is that it must have real data labels, so it can only be used in supervised scenarios, and cannot be integrated into unsupervised or semi-supervised learning.

In addition, mixup has only verified its effectiveness in the image field, and it cannot be applied in the video field

The existence of these difficulties has led to the exploration of video data enhancement for a long time only in simple rotation and color dithering.

Images