This paper addresses the class imbalance problem in real-world object detection and classification tasks, where some classes are over-represented while others are under-represented. The authors propose a cost-sensitive deep neural network that can automatically learn robust feature representations for both majority and minority classes. During training, the network jointly optimizes class-dependent costs and neural network parameters. The approach is applicable to both binary and multi-class problems without modification and does not alter the original data distribution, reducing computational costs. The proposed method is evaluated on six major image classification datasets, showing significant improvements over baseline algorithms and state-of-the-art techniques. The results demonstrate the superior performance of the proposed method in handling class imbalance.This paper addresses the class imbalance problem in real-world object detection and classification tasks, where some classes are over-represented while others are under-represented. The authors propose a cost-sensitive deep neural network that can automatically learn robust feature representations for both majority and minority classes. During training, the network jointly optimizes class-dependent costs and neural network parameters. The approach is applicable to both binary and multi-class problems without modification and does not alter the original data distribution, reducing computational costs. The proposed method is evaluated on six major image classification datasets, showing significant improvements over baseline algorithms and state-of-the-art techniques. The results demonstrate the superior performance of the proposed method in handling class imbalance.