The challenge of this project was to automate the detection of air conditioning cooling towers in a region, using a time series database of orthophotos and a training dataset with geolocated annotated features. The images used, captured biannually since 2004, contained varying resolution over the years, making the analysis more complicated. The task was not only to develop a deep learning model capable of recognizing and accurately pinpointing the cooling towers from these orthophotos, but also to optimize the dataset based on the chosen model. A substantial amount of time and resources would traditionally be required for manual detection and analysis, leading to the need for a more efficient and precise solution.
Randbee Consultants tackled this challenge by leveraging the power of YOLO, a state-of-the-art object detection model. The team first carried out preprocessing of the orthophotos to standardize them, regardless of the changes in resolution over the years. They then trained the model on the provided dataset, which had been optimized for this specific model. The deep learning model developed was capable of identifying and pinpointing the cooling towers automatically, greatly improving the detection accuracy. This innovative solution saved significant time and resources, making the process of detecting air conditioning cooling towers in a given region efficient and highly accurate.
Copyright © 2023 Randbee Consultants