This study deals with the development of an early monitoring and alert system for crack detection based on systematic capture of images from ore leach dumps surfaces. The system developed includes the capture, processing and visualization of images aiding in early detection of crack formation and/or settling using a neural network as our basis. An existing base model for general crack detection was adapted to the particular setting of our problem and the operating conditions of an ore leach dump using transfer learning. The results obtained indicate that the model tailored for the case of an ore leach dump improves performance compared with existing general models for cracks detection. The monitoring system based on the proposed model has been developed in collaboration with Minera Antucoya. A first implementation of the system has given positive results in the early detection of cracks on ore leach dumps.