Adaptive Weighted 3D Object Image Inference Model Based on Image Complexity

Abstract

The research on product style classification based on CNN is very active, but the data used to train CNN(Convolutional Neural Networks) are often single-view images of 3D objects, which will lead to the loss of unpredictable object feature information and does not match the real scene. It reduces the quality of the model training. This paper proposes an adaptive weighted CNN model based on image complexity. Feature extraction is performed on images of 3D objects from different perspectives through convolutional neural networks, and the final classification result is obtained by weighting based on image complexity. The 3D object discrimination model in this paper is more in line with the cognitive process of the audience, and can improve the quality of style inference of 3D objects.

Keywords: image complexity, convolutional neural networks, cognitive inference, three-dimensional model recognition

DOI: 10.54941/ahfe1001832