performance analysis of segmentation models of computer vision and its integration with ros to pick up rocks and minerals using a robotic arm in mars
- Paper number
IAC-24,A3,IP,78,x86430
- Author
Mr. Belyeud Prado, Universidad Nacional de Ingeniería (Lima, Perù), Peru
- Coauthor
Mr. Paulo Cesar Romero Aguilar, Universidad Nacional de Ingeniería (Lima, Perù), Peru
- Coauthor
Mr. Diego Martin Arroyo Villanueva, Universidad Nacional de Ingeniería (Lima, Perù), Peru
- Coauthor
Mr. Brayam Donayre Farfan, Universidad Nacional de Ingenieria, Peru, Peru
- Coauthor
Mr. Freddy Dick Salazar Valverde, Universidad Nacional de Ingeniería (Lima, Perù), Peru
- Year
2024
- Abstract
In this paper, it focuses on a complete evaluation study of performances, in a Jetson Nvidia, when integrating Computer Vision segmentation models in a robotic arm using ROS. The main objective is to determine which model is more efficient to improve performance and speed in a robotic arm for future explorations on Mars. The algorithms to be evaluated are the latest version of YOLO (YOLOV9), DinoV2 and Fast-SAM. These models were trained with several images of rocks and minerals characteristic of Mars and the number of images was increased with Data Augmentation techniques to have better results. The first results indicate a significant improvement in performance and response time of the robotic arm since it segments the object to be picked up much faster. These results offer a valuable resource to consider for future explorations on Mars.
- Abstract document
- Manuscript document
(absent)