ROS Image Script

[sjcantor]

Current State

The image script currently reads in new images from the robot, converts it to a numpy array, displays it using openCV and saves it to a jpeg.

Next Steps

• Need to figure out how to display this as a video instead of the current implementation, where you have to hold spacebar to replace the current image with a new one.
• The image should be published to a new topic/node, and a new script will subscribe to that topic to perform object recognition

[sjcantor]

This issue has been fixed for the most part. Live feeds of the image with and without bounding boxes can be viewed in rviz or through iage_transport, but the stream with bounding boxes has a noticeable delay. Surprisingly, this delay is not bottle-necked by the deep learning model, but instead by the matplotlib figure and publishing node. Will look into more efficient ways of plotting boxes.