![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
I'm back in Vancouver now. My big adventure (unless I'm able to do Burning Man) for 2009 is over. I still need to do a proper writeup of the inauguration, but not tonight. I'm just too tired.
Instead, let me distract you with an animation of a series of image segmentations I recently finished doing for my research. These are every 5 frames of a short section of stereo video I captured moving through the lab. They will serve as the ground truth for training a classifier for automatic segmentation based on the motion of 3D pointclouds derived from the stereo video data.
This is just a rough test for getting the system working. The semantic categories I used here won't be the final ones, and I'll do it in higher resolution for the real data. Also, I'll need to use a real robot performing random exploration, instead of pushing an office chair around.
Instead, let me distract you with an animation of a series of image segmentations I recently finished doing for my research. These are every 5 frames of a short section of stereo video I captured moving through the lab. They will serve as the ground truth for training a classifier for automatic segmentation based on the motion of 3D pointclouds derived from the stereo video data.
This is just a rough test for getting the system working. The semantic categories I used here won't be the final ones, and I'll do it in higher resolution for the real data. Also, I'll need to use a real robot performing random exploration, instead of pushing an office chair around.