Fusion of evidential CNN classifiers for image classification
This work addresses image classification accuracy for computer vision applications, but it is incremental as it builds on existing CNN and fusion methods.
The paper tackles image classification by fusing multiple pre-trained CNN classifiers using belief functions and Dempster's rule, achieving improved performance as demonstrated on three benchmark databases.
We propose an information-fusion approach based on belief functions to combine convolutional neural networks. In this approach, several pre-trained DS-based CNN architectures extract features from input images and convert them into mass functions on different frames of discernment. A fusion module then aggregates these mass functions using Dempster's rule. An end-to-end learning procedure allows us to fine-tune the overall architecture using a learning set with soft labels, which further improves the classification performance. The effectiveness of this approach is demonstrated experimentally using three benchmark databases.