Clothing identification method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a clothing identification method and device, equipment and a storage medium. The method comprises the following steps: extracting key frames at multiple different moments from a to-be-identified video; carrying out costume area detection on the key frame; for the detected target clothes area, extracting features of the target clothes area, the features of the target clothes area at least comprising image features; on the basis of the features of the target clothes areas, clustering the target clothes areas to obtain at least one type of target clothes areas; and for each type of target clothes area, fusing the image features of each target clothes area in the type of target clothes area into a target image feature, and performing clothes identification based on the target image feature to obtain an identification result of the type of target clothes area. Therefore, the accuracy of clothes identification is improved.

Description

technical field [0001] The present invention relates to the technical field of data processing, in particular to a clothing identification method, device, equipment and storage medium. Background technique [0002] Clothing is an indispensable article in people's daily life. With the emergence of online shopping platforms, people often buy clothing on online shopping platforms, which greatly facilitates people's lives. In related technologies, in order to further provide convenience for people, clothing appearing in a video can be identified through clothing recognition technology to provide people with relevant clothing information. The existing clothing recognition technology is still in the development stage, and the recognition accuracy rate is still low, which makes it difficult to meet the recognition requirements of practical application scenarios. For example, the maximum accuracy rate of most clothing recognition technologies based on open source data sets is lower...

AI Technical Summary

Problems solved by technology

The existing clothing recognition technology is still in the development stage, and the recognition accuracy rate is still low, which makes it difficult to meet the recognition requirements of practical application scenarios. For example, the maximum accuracy rate of most clothing recognition technologies based on open source data sets is lower than 70%. It is difficult to meet the recognition requirement that the accuracy rate required by the actual application scenario is higher than 90%.

In particular, the following recognition problems are prone to occur: since the clothing is a non-rigid object, the difference in the posture of the characters is likely to cause the occlusion and incomplete exposure of the clothing area, which may lead to t 1 Moment (such as the moment when the front of the garment is shown) is related to t 2 Moments (such as the moment when the back of the garment is shown) the same garment at two different times are identified as different garments, or t 3 Moments (such as showing the back of a plain white T-shirt) are associated with t 4 Moments (for example, showing a T-shirt with a large pattern on the front and chest, but the back of a pure white T-shirt) two different clothing appearing at two different moments are recognized as the same clothing problem, which is the problem of clothing recognition in the video Severe challenge

Images