Model weights and configs for yolo_ros. Detection and segmentation YOLO models are supported.
config: Configuration files for different competitions.models: Pre-trained YOLO models in PyTorch.ptformat.launch: A launch file for this package. Accepts arguments:rgbd_ids=(space separated camera names, e.g. d455_front)config=(name of config file without .yaml suffix)
Use yolo-gym to train a detection or segmentation model and copy the PyTorch weights over here to the models directory. Then create a new config file in the config directory. Remember to update the parameters, in particular: model, class_names, class_radii. For segmentation models, instance masks are used internally for depth/size estimation when available; Humble vision_msgs/Detection2D does not publish masks in /detections.