Identifying plant diseases at early stages is crucial as they directly impact global food security. Traditionally, identifying disease has relied on manual inspection by plant pathologists, a time-consuming and laborious process. Therefore, an automated plant disease identification model is needed. This study presents a Convolutional Block Attention Module based Lite DenseNet (CBAM-LDNet) architecture to identify plant disease by incorporating the Lite DenseNet architecture enhanced by the Convolution Block Attention Module (CBAM). The proposed approach utilizes dense blocks for efficient feature extraction and CBAM for adaptive attention. Additionally, focal loss is utilized to handle imbalanced datasets. The CBAM-LDNet significantly reduces computational costs while maintaining high accuracy. The performance is compared with traditional techniques and other state-of-the-art methods, including pre-trained convolutional neural networks. Its ability to effectively focus on subtle, complex features and handle class imbalance contributes to its superior performance on certain datasets. Extensive experiments show that the CBAM-LDNet achieves superior accuracy and robustness and highlights effectiveness in real-time plant identification tasks.