Extract Better Features
Currently non-morphological features are very simple, such as extracting the RGB or hue value at the center of the sensor patch.
In the short term, we would like to extract richer features, such as using HTM's spatial-pooler or Local Binary Patterns for visual features, or processing depth information within a patch to approximate tactile texture.
In the longer-term, given the "sub-cortical" nature of this sensory processing, we might also consider neural-network based feature extraction, such as shallow convolutional neural networks, however please see our FAQ on why Monty does not currently use deep learning.
Note that regardless of the approach taken, features should be rotation invariant. For example, a textured pattern should be detected regardless of the sensor's orientation, and the representation of that texture should not be affected by the sensor's orientation.
Help Us Make This Page Better
All our docs are open-source. If something is wrong or unclear, submit a PR to fix it!
Updated 14 days ago