Clothing target detection method based on Faster R-CNN method

Abstract

The invention discloses a clothing target detection method based on a Faster R-CNN method. The method comprises the following steps: step 1, preparing a data set; step 2, constructing a Faster R-CNN algorithm framework; 3, resetting the size of the bounding box by using a K-Means clustering algorithm; 4, training the clothing detection model by using the training set A; and step 5, testing the model performance by using the remaining test set. According to the method, the retrieval speed is increased. According to the method, a K-Means clustering algorithm is introduced, so that the size of aregion frame is closer to the size of actual clothes, and while the speed is increased, the recognition precision is also improved to a certain extent.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to a clothing target detection method based on the Faster R-CNN method and a K-Means clustering algorithm. Background technique [0002] Object detection technology has always been one of the topics that scholars around the world have paid much attention to, and deep learning methods are a good weapon for studying computer vision. It has been very effectively applied to visual recognition problems, such as large-scale vision Identifying the ImageNet challenge, and PAS-CAL, object classification and detection tasks on the VOC dataset. Now, with the rise of the e-commerce industry, the demand for target detection technology in clothing has also become stronger. For example, shopping websites detect and classify the products that users like, so as to calculate the types of clothing that users like, and use this to make personalized recommendations to users, which gre...

AI Technical Summary

Problems solved by technology

[0004] In the target detection method based on regression calculation, the core of the YOLO method is to solve a regression problem, and use an independent network to discriminate and output the results. Obviously this makes the detection speed of YOLO very fast, but it shows that YOLO is very close to the distance or The detection effect of objects with overlapping parts is not good. For this reason, the SSD method combines the region proposal network structure to train a new model to make up for its detection accuracy problem.

[0005] Comparing the two types of algorithms, the R-CNN series algorithms have higher accuracy for target detection, but the detection speed is not fast enough, while the YOLO series is faster than Faster R-CNN although the accuracy is lower

Images