Performance Measurement of Classification Algorithms for Aerial Image Registration

The random forest algorithm is a popular machine learning technique that is widely used for classification and regression tasks. Although it is known for its high accuracy and robustness, one of the main challenges associated with the random forest algorithm is its long execution time, particularly when dealing with large datasets. Therefore, several methods have been existed to reduce the execution time of the random forest algorithm, including optimization of hyperparameters, feature selection method and parallelization techniques. In order to do that, the Random Forest (RF) parameters have been proposed, and the experiments for tuning four RF parameters are illustrated based on CPU/GPU with PC and Nvidia Jetson board. In this paper, the results of execution time with Windows and Linux operating systems are presented. NVIDIA’s Jetson platform offers great potential for embedded machine learning, aiming to strike a harmonious balance between the objectives of high accuracy, throughput and high performance. We discussed the weakness and strength for each params, and provided insights into their implementation and performance. The results of Jetson Nano board shown that the proposed methods can significantly reducing the execution time of RF params. All the coding steps are available at https://github.com/HayderMosaMerza/image_registration .