PCLNet learns multi-frame optical flow unsupervisedly via pyramid ConvLSTM and frame reconstruction, decoupling motion features from flow representation and achieving comparable action recognition performance.
Deepflow: Large displacement optical flow with deep matching,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Unsupervised Learning for Optical Flow Estimation Using Pyramid Convolution LSTM
PCLNet learns multi-frame optical flow unsupervisedly via pyramid ConvLSTM and frame reconstruction, decoupling motion features from flow representation and achieving comparable action recognition performance.