Weak supervision specific behavior recognition system based on multi-level labels

Abstract

The invention provides a weak supervision specific behavior recognition system based on multi-level labels. The weak supervision specific behavior recognition system comprises a detection information acquisition unit, an information storage unit and a specific behavior recognition unit; the detection information acquisition unit is used for acquiring video data through a camera device, preprocessing acquired to-be-identified specific behavior video image data and then storing the data to the information storage unit; the information storage unit is used for storing video data and operation parameters collected by the system, and the information storage unit comprises a label knowledge base, a training set database and a video cache module; the specific behavior recognition unit is used for recognizing the preprocessed image data, extracting image features through a convolutional neural network based on weak supervision, and combining with multiple sub-networks to obtain a classification network model to perform specific behavior classification recognition; the invention can perform parallel operation on the input image and is high in calculation efficiency.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a weakly supervised specific behavior recognition system based on multi-level labels. Background technique [0002] In strong supervised learning, it is generally necessary to mark images at the target level or semantic level to obtain supervisory information. When processing a large number of images, there is a problem of difficulty in marking. The image-level labels used in weakly supervised learning only mark which categories of objects exist in the image, and there is no label for the position information of the existing objects in the image. However, under the condition of weakly supervised learning, it is difficult for the learning model to directly locate the region of the target in the image. The existing weakly supervised learning methods roughly include the following two categories: one is to obtain candidate frames through targeted algorithms, etc., and throu...

AI Technical Summary

Problems solved by technology

The former is expensive to obtain the target candidate frame, and there are a lot of redundant calculations in the classification process; while the latter's time-series progressive method makes it impossible to perform efficient parallel operations, and the calculation efficiency is low

Images