Details: This package is based on ERFNet, a semantic segmentation deep learning
network architecture. We have packaged together a quick and simple way to train
this model on two classes for fast and accurate object segmentation. This
package is designed to work on the OVC platform and we have achieved accurate
object detection results at 10Hz at half the original camera resolution.
Details: For our robotics related applications, we use a pair of wide-angle
lenses for our cameras. This package performs a fisheye rectification of input
images, either in the vertically concatenated stacked format or as synchronized
separate image streams.
Details: A ROS node of the Semi-Global Matching algorithm implemented originally by
Hernandez et. al with a few modifications and a ROS front-end. Performs dense
stereo depth estimation, with point cloud, disparity image and regular 8bit
disparity image topics.