Many bugs have been fixed since v1.0, the main modifications are as follows:
- Face recognition is now available, we have fine-tuned the VGG_FACE model on our own face dataset and use it to recognize our teammates, till now the training process is off-line.
- Added Vision Action to deal with interactions with other programs of the robot, with the help of smach.
- Image topics and point cloud topics have been reduced to cut down the consumption of bandwidth, now the vision system can run around 20Hz on multiple machines configuration.