The apple tree, which is extensively consumed globally, has gained significant attention across various research domains. It is crucial to prevent and manage diseases in an apple tree in order to increase its yield and meet its ever-increasing demand. Researchers have developed a variety of ways for identifying plant diseases using the techniques of machine learning and deep learning methodologies. Since deep learning models have an enormous set of parameters, they are not suitable for resource-constrained environments. To address this problem, A Novel Efficient Compressed Convolutional Neural Network (ECCNN) is proposed to achieve an efficient and accurate classification of apple leaf disease in this study. The proposed network architecture begins by utilizing the initial two VGG16 blocks to perform efficient feature extraction. Subsequently, batch normalization is employed after two blocks of the VGG16. The output of the batch normalization is subjected to a convolutional block to reduce the number of parameters. Finally, a dense and softmax layer is used to accurately classify the apple leaf disease. A number of comparative experiments were carried out using existing cutting-edge convolutional neural network systems (CNN) such as VGG16, Xception, Inception, EfficientNetB0, MobileNet V2, and ShuffleNetV2. Our findings demonstrate that ECCNN is evaluated on an apple leaf dataset and achieves an accuracy of 98.85%. Extensive experiments show that the model outperformed and had significantly lower parameters than existing cutting-edge methods.