YOLO Object Detectors for Robotics -- a Comparative Study
For robotics practitioners, this work offers a comparative analysis of YOLO models under realistic distortions, but it is incremental as it applies existing detectors to a specific domain without novel methods.
This study evaluates multiple YOLO object detector versions for detecting objects in robot workspaces, using custom and COCO2017 datasets with distorted images. The results provide guidance on selecting suitable YOLO versions for robotic vision tasks.
YOLO object detectors recently became a key component of vision systems in many domains. The family of available YOLO models consists of multiple versions, each in various variants. The research reported in this paper aims to validate the applicability of members of this family to detect objects located within the robot workspace. In our experiments, we used our custom dataset and the COCO2017 dataset. To test the robustness of investigated detectors, the images of these datasets were subject to distortions. The results of our experiments, including variations of training/testing configurations and models, may support the choice of the appropriate YOLO version for robotic vision tasks.