Mimetics as a Learning Basis - Part 2
Wrapping up the exploration of mimetics in learning from Part 1, the causal model is updated to employ the following continuous data processing pipeline:
- Low-level processing/perception
- Consume video stream, convert to 2D keypoints, then 3D skeletal, view-invariant form
- Learning
- Infer with existing signatures/sequences and score results
- Cluster and store new signatures/sequences
The concepts explored here are also applicable to the memory system of the human mind in that memories are (re)constructed from encodings, not recorded and then played back. In addition to remembering, this allows individuals to imagine what-if and future scenarios.
The following video clip demonstrates mimicry of an observed subject (left side) constructed from visual encodings (right side). With some additional semantic encodings and storage, this becomes a memory.