darknet_ros

Leverages CUDA for high-speed inference on Nvidia GPUs, enabling real-time object detection crucial for robotics applications, as highlighted in the README with performance comparisons.

Publishes bounding boxes, detection images, and object counts as standard ROS topics and actions, making it easy to integrate into existing ROS pipelines for navigation or manipulation tasks.

Supports training and deployment of user-defined YOLO models, allowing adaptation to specialized detection needs, with configurable weights and cfg files as described in the installation section.

Includes pre-trained weights for COCO and VOC datasets, providing immediate functionality for common object classes without additional training effort.