Cracks in buildings can significantly compromise structural integrity, posing safety hazards and requiring timely intervention. Manual crack detection methods are labor-intensive and prone to subjectivity, motivating the exploration of automated approaches. This paper proposes a robust building crack detection method by harnessing the capabilities of convolutional neural network. The focus of this research is to investigate a crack identification method that makes use of completely convolutional block detection in order to overcome the difficulty of fracture detection. The ResNet fully convolutional neural network architecture is integrated with the idea of block detection in the method. It gets over the limitations of regular block detection receptive fields by using a local discriminating technique. This method outperforms the standard ResNet picture classification system in terms of generalizability and detection performance, especially when dealing with minor crack cases. Our novel approach addresses the challenges associated with crack detection through an end-to-end deep learning framework that learns discriminative features directly from raw building images. Extensive experimentation has provided concrete evidence to support the efficacy of this method.