Abstract:
Real-world image classification problems often have class imbalance, with some classes having abundant data and others not. In this case, classifier representations are biased toward the majority classes and learning proper features is difficult, resulting in poor performance. Many algorithm-level methods explicitly prioritize minority classes based on data distribution to eliminate this biased feature representation. This article proposes data-driven deep attention-based imbalanced image classification (DAIIC) to automatically pay more attention to minority classes. An attention network and a novel attention augmented logistic regression function are used to encapsulate as many minority class features as possible into the discriminative feature learning process by assigning class attention jointly in both prediction and feature spaces. The object function allows DAIIC to automatically learn class misclassification costs. Then, using the designed attention networks, the learned misclassification costs can guide training to learn more discriminative features. The method works for various networks and data sets. Experimental results on single-label and multilabel imbalanced image classification data sets show that the proposed method has good generalizability and outperforms several state-of-the-art methods.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
