Land Cover and Land Use (LCLU) classification is critical in remote sensing, especially when analyzing hyperspectral and RGB images. LCLU classification with these image types is inherently challenging due to the limited availability of training samples. In this paper, a deep learning (DL) approach is developed to address this issue. For hyperspectral image analysis, multiple models were constructed, including 2D and 3D Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Units, and bidirectional LSTM models. Pre-trained CNN architectures such as VGG16, VGG19, ResNet50, ResNet50 V2, and MobileNet were employed for RGB image analysis. These models were applied to five datasets: Pavia University (PU), Kennedy Space Center (KSC), and Indian Pines (IP) for hyperspectral images, and Eurosat (ES) and UC Merced (UCM) for RGB images. The proposed models achieved high accuracy, with bidirectional LSTM reaching 99.2% accuracy on the PU dataset, 2D CNN attaining 99.6% on the KSC dataset, and 2D CNN achieving 100% on the IP dataset. For RGB image analysis, VGG16 achieved 97.81% accuracy on the ES dataset, while ResNet50 reached 98.6% on the UCM dataset. Comparative analysis with the other nine different DL models indicates that the proposed approach demonstrates superior accuracy and performance across all tested datasets.