Buildings usually develop cracks due to its inability to withstand stress in the components over a period of time or due to natural calamities such as earthquakes. Detecting cracks during the early stage helps in longing the lifespan of the building and avoid catastrophic collapses. Detecting cracks is done as visual inspection by inspectors, which are usually time consuming, erroneous, dangerous and expensive. In order to monitor all the elements of the high raised buildings, ensure the safety of the inspector and to revise the crack images in various angles, this application can be used. It detects the crack from images captured through drones using Machine Learning algorithms and is trained with a set of pre-defined pictures (presence of crack - positive/ absence of crack - negative) to identify cracks. Using image pre-processing techniques such as Sobel Edge Detection, the undesired details from the image are eliminated, which results in the increase of performance and accuracy. It is developed using python with imported libraries from TensorFlow, NumPy, Pandas, MatPlotLib, cv2, and PIL. The model is trained and tested with the help of Convolutional Neural Network (CNN). The reason behind choosing CNN is that it automatically detects important features without human supervision and there is no need to specify the filters to be used. CNN considers context information in a tiny environment and this helps in achieving better accuracy. Finally, Images are stitched together to visualise the entire crack. The application is based on a model which provides 90.67% accuracy and 89.8% valaccuracy.