Research on Painting Image Classification Based on Transfer Learning and Feature Fusion

Abstract

In order to effectively solve the problems of high error rate, long time consuming, and low accuracy of feature extraction in current painting image classification methods, a painting image classification method based on transfer learning and feature fusion was proposed. The global characteristics of the painting picture, such as color, texture, and form, are extracted. The SIFT method is used to extract the painting’s local features, and the global and local characteristics are normalized and merged. The painting images are preliminarily classified using the result of feature fusion, the deterministic and nondeterministic samples are divided, and the estimated Gaussian model parameters are transferred to the target domain via a transfer learning algorithm to alter the distribution of nondeterministic samples, completing the painting image classification. Experimental results show that the proposed method has a low error rate and low feature extraction time and a high accuracy rate of painting image classification.

Revolutionizing Waste Management: A Cutting-edge pyTorch Model for Waste Classification and Prediction, Coupled with a User-friendly Recycling Recommendation Application Built with Tkinter

“…When we compare our results to those of other relevant research, we find that our model's performance is on par with or better than that of earlier efforts in garbage categorization. For example, Thung and Yang reached an 87% accuracy using a ResNet-50 and SVM technique [1], while another study utilizing a unique trash image recognition model revealed a high recall rate and accuracy [2]. Furthermore, our model outperformed other techniques, including Xception_CutLayer and InceptionResNetV2_CutLayer, which had accuracies of 89.72% and 85.77%, respectively [3].…”

“…The dataset has been expanded through data augmentation; the resulting dataset contains 41,650 garbage images. The results of the experiments demonstrate that the suggested model has good convergence, a high recall rate and accuracy, and a short recognition time when compared to other models [2]. Six distinct waste materials were used in the training and testing of the suggested techniques and AI-based deep learning techniques.…”

“…Thung and Yang's Trash Image Dataset yielded an accuracy of 87% when ResNet-50 and SVM were combined [1]. The true value of the GCNet, which is a combination of Vision Transformer, DenseNet, and EfficientNetv2, is unknown, but it showed excellent accuracy on an augmented garbage image dataset [2]. Inception and Xception_CutLayer with accuracy rates of 89.72% and 85.77%, respectively, on a dataset comprising six different waste materials, ResNetV2_CutLayer performed better than other methods [3].…”

“…Nevertheless, by utilizing PyTorch's deep learning capabilities, the suggested model seeks to close current research gaps. The utilization of diverse neural network architectures for garbage image recognition is crucial, as evidenced by the potential improvements in convergence, recall rate, and accuracy suggested by the combination of model fusion and transfer learning, as demonstrated in GCNet[2]. By presenting a PyTorch-based methodology, our study adds to this field of study and may have advantages over other models in terms of recognition time, accuracy, and convergence speed[2].…”

Practical application of modern extracurricular painting landscape art to promote environmental protection

“…With the development of contemporary painting, modern landscape art pays more attention to fun. Sex and aesthetics, this change is also inseparable from the development of painting art [14].…”