Some notes here.
Video feed from camera is obtained using openRTSP. The feed is saved to a file
which will then be read by the vision_node to perform object detection.
Object detection is done by performing regular image classification with a CNN over a sliding window. The neural network framework used is darknet.
https://gist.github.com/randolphwong/fc4077abc8c51ec5f0662d359869337e
rosrun amdc vision_node data/cnn_configuration data/cnn_weights videofeed.mp4