Small sample commodity image classification method, a device, equipment and storage medium

Abstract

The invention relates to a small sample commodity image classification method, a device, equipment and a storage medium. The method comprises the following steps: collecting a target commodity image; inputting the target commodity image and the support set image to a trained commodity classification model, obtaining the similarity between the commodity image and each image in the support set image, wherein the commodity classification model comprises an embedding module and a measurement module, the embedding module comprises a first convolution layer and a first ECA layer which are connected with each other, the measurement module comprises a second convolution layer and a full connection layer which are connected with each other; and obtaining the category of the target commodity according to the maximum similarity. According to the method, the ECA module is integrated into the embedding module, so that a more accurate feature map can be extracted, and subsequent measurement module classification is more effective. Meanwhile, the measurement module integrated with the ECA module can utilize the relationship between different types of image features while comparing the feature similarity, so that the classification accuracy is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a small-sample product image classification method, device, equipment and storage medium. Background technique [0002] With the advent of the big data era, deep learning models have achieved advanced results in tasks such as image classification and text classification. But the success of deep learning models depends largely on a large amount of training data. However, in real-world scenarios, some categories have only a small amount of data or a small amount of labeled data, and labeling unlabeled data will consume a lot of time and manpower. [0003] Just like in the retail industry, in the classification of retail product images, due to the large number of product categories, similar features, and few labeled samples, the traditional deep learning model is limited. At present, there are many methods to solve the small sample classification task, but there i...

AI Technical Summary

Problems solved by technology

Although the base learner can adapt to changes, meta-learning also faces the problems of task bias and overfitting

Moreover, in the meta-learning method, complex memory structures are often used, and optimization and training are relatively difficult.

Images