How do robots recognize different shapes and dimensions?

The answer is - data labelling! Video annotation with cuboids allows computer vision algorithms to add depth perception to a 2d photo

robotsrecognizedifferent