Urbanization detection in remotely sensed images is a very important task as it helps us to manage organized city development, resource planning, disaster impact assessment, and various other urban management challenges. In this research, we introduce EnsembleNet, a new method for urbanization observation via building change detection on the LEVIR-CD dataset. We make use of ResNet18 to learn multi-level spatial features with an added Squeeze and Excitation block (SE) for channel attention. To further enhance detection, we propose a Spatial and Channel Attention (SCA) module. We compare two image differencing methods within a difference module, and then we apply custom weights to the models based on their performance to enhance change detection and combine their performance. The ensemble model yields an overall accuracy of 98.71% and recall of 93.43%, successfully detecting true changes and reducing pseudo changes.