Sample image processing method and device, electronic equipment and storage medium

Abstract

The invention provides a sample image processing method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a scene model and multiple object models; stacking a plurality of object models in the scene model; stacking the three-dimensional model of the target object at the corresponding position and posture in the scene model according to the multiplepositions and postures of the target object to obtain multiple virtual scenes containing the target object; and generating a virtual scene image and a corresponding label according to the configured shooting range of the virtual camera to obtain a sample image set. According to the scheme, the sample image acquisition efficiency can be improved, manual annotation is not needed, the workload of manually collecting training samples and annotating is reduced, and a problem that manual annotation is prone to errors is solved so that the accuracy of an algorithm model obtained on the basis of sample image training is improved.

Description

Technical field[0001]This application relates to the field of image processing technology, and in particular to a method and device for processing sample images, electronic equipment, and storage media.Background technique[0002]In the field of commodity retail, channel monitoring based on deep learning solutions has been widely adopted. In the existing solution, the customer inputs the image of the retail product, and after being identified by the CNN detection model, the customer obtains the category of the product contained in the image and the position in the image.[0003]The current technical solutions have low efficiency and accuracy when dealing with new scenarios. Up-to-date means that when a customer launches a new retail product, the deep learning model needs to be updated simultaneously so that the model has the ability to recognize the new product. When the current technical solution is updated, it is necessary to complete the three steps of data collection, data labeling,...

AI Technical Summary

Problems solved by technology

When the current technical solution is updated, three steps need to be completed: data collection, data labeling, and retraining. Data collection and labeling are all manual operations, resulting in low efficiency

Since the data collection of new products is generally carried out in the laboratory scene, the data obtained in this way is different from the online data in terms of data diversity, which leads to the low recognition accuracy of the model trained by self-collected data for online new products

Images