<p>The pineapple is a tropical fruit that often causes losses to farmers due to poor management of diseases. Early detection of diseases, followed by appropriate remedial actions in pineapple farming, can significantly increase yield rates and improve fruit quality. Since the quality of pineapple fruit is crucial for farmers to gain profits, addressing this issue is essential. Advanced deep learning approaches for image detection and classification are suitable solutions to tackle this problem. Recently, channel and spatial attention-based multi-layered feature fusion models have gained significant attention in the research community. In this context, this research focuses on the early detection and classification of pineapple fruit diseases (crown rot, fruit fasciation, fruit rot, mealybug wilt, and no disease) using a Channel and Spatial Attention mechanism followed by Adaptive Channel Reduction-based multi-layered feature fusion with the DenseNet121 model. Transfer learning models such as DenseNet121, InceptionV3, VGG-16, Xception, ConvNeXtBase, and ResNeXt50_32x4d were applied to the pineapple fruit disease dataset. Among these, DenseNet121 demonstrated the highest accuracy. Integrating DenseNet121 with Channel-Spatial Attention and Adaptive Channel Reduction (CSAACR) further enhanced the classification performance, achieving an accuracy of 95.1% in classifying pineapple fruit diseases. The integration of this proposed CSA-ACR-DenseNet121 model with a mobile application could help farmers in early disease detection and determining appropriate remedial actions, ultimately leading to higher-quality and greater pineapple yields.</p>