Scalable Full Flow with Learned Binary Descriptors

Gottfried Munda, Alexander Shekhovtsov, Patrick Knöbelreiter, Thomas Pock

Research output: Contribution to conferencePaperpeer-review


We propose a method for large displacement optical flow in
which local matching costs are learned by a convolutional neural network
(CNN) and a smoothness prior is imposed by a conditional random
field (CRF). We tackle the computation- and memory-intensive operations
on the 4D cost volume by a min-projection which reduces memory
complexity from quadratic to linear and binary descriptors for efficient
matching. This enables evaluation of the cost on the fly and allows to
perform learning and CRF inference on high resolution images without
ever storing the 4D cost volume. To address the problem of learning binary
descriptors we propose a new hybrid learning scheme. In contrast
to current state of the art approaches for learning binary CNNs we can
compute the exact non-zero gradient within our model. We compare several
methods for training binary descriptors and show results on public
available benchmarks.
Original languageEnglish
Publication statusPublished - 13 Sept 2017
Event39th German Conference on Pattern Recognition - Basel, Switzerland
Duration: 13 Sept 201615 Sept 2017


Conference39th German Conference on Pattern Recognition
Abbreviated title GCPR 2017
Internet address
  • GCPR Honorable Mention

    Munda, Gottfried (Recipient), Shekhovtsov, Oleksandr (Recipient), Knöbelreiter, Patrick (Recipient) & Pock, Thomas (Recipient), 15 Sept 2017

    Prize: Prizes / Medals / Awards

Cite this