TY  - BOOK
AB  - In this paper we present an incremental approach to learning generative models of object manipulation actions as HMMs over qualitative relations between two objects. We compare the incremental approach against a traditional batch training baseline and show that the resulting qualitative action models are capable of one-shot learning after just one seen example while displaying good generalization behavior as more data becomes available.
DA  - 2015
KW  - qualitative trajectory calculus
KW  - action recognition
KW  - model merging
KW  - hidden markov models
KW  - online classification
LA  - eng
PY  - 2015
TI  - Incremental learning of action models as HMMs over qualitative trajectory representations
UR  - https://nbn-resolving.org/urn:nbn:de:0070-pub-27754143
Y2  - 2024-11-22T02:09:27
ER  -