Skip to main content
. Author manuscript; available in PMC: 2022 Dec 1.
Published in final edited form as: Med Image Anal. 2021 Sep 3;74:102224. doi: 10.1016/j.media.2021.102224

Fig. 1:

Fig. 1:

Overview of our phase recognition system. The system takes depth videos as input, and the phase is predicted using 32-second-frame inputs to the short-term module and 320-second-frame inputs to RLO. The outputs of the short-term and long-term modules are fused for making the final prediction.