What can we expect from a V1-MT feedforward architecture for optical flow estimation?

作者：

Highlights：

• We show how a V1-MT neural model can be adapted to handle real sequences.

• For the first time, a neural model is benchmarked on a Middlebury dataset.

• We share our code to encourage research in bio-inspired scalable vision algorithms.

摘要

Highlights•We show how a V1-MT neural model can be adapted to handle real sequences.•For the first time, a neural model is benchmarked on a Middlebury dataset.•We share our code to encourage research in bio-inspired scalable vision algorithms.

论文关键词：Optical flow,Spatio-temporal filters,Motion energy,V1,MT,Benchmarking

论文评审过程：Available online 2 May 2015, Version of Record 18 November 2015.

论文官网地址：https://doi.org/10.1016/j.image.2015.04.006