Land Use and Land Cover (LULC) classification of remote sensing images is a challenging task. Land Use represents the land utilized by humans, e.g., agricultural land, urban areas, parks, etc. Land cover represents the physical material present on Earth’s surface. This study presents a detailed comparative analysis of three advanced deep learning models, namely fuzzy convolutional neural networks, fuzzy CNN generative adversarial networks integrated with GAN-CNN, and a Hybrid CNN-LSTM Architecture for LULC classification. The models are tested on the UC Merced Land Use Dataset. Each model is examined in terms of its architectural design, training methodology, and evaluation based on metrics such as accuracy, area under the curve (AUC), and confusion matrix. Among these three models experimented, the maximum accuracy is obtained for the GAN-CNN model ( \(93.65\%\) ), followed by the Fuzzy CNN with \(90.79\%\) and the CNN-LSTM model with \(83.81\%\) . The findings highlight the potential of the three mentioned models and highlight the mechanism to overcome the challenges of applying different deep learning approaches to remote sensing tasks, offering insights into their respective strengths and suitability for LULC classification.