DeepVO

DeepVO for Visual Odometry Implementation using Tensorflow.

Install

Build Custom Layers

Custom layers are implemented, so you should build them before you use.

See This.

Dependencies

• For Test
  • PySophus : https://github.com/arntanguy/PySophus

References

Papers & Lectures

• DeepVO: Towards End-to-End Visual Odometry with Deep Recurrent Convolutional Neural Networks
• (ESP-VO) End-to-End, Sequence-to-Sequence Probabilistic Visual Odometry through Deep Neural Networks
• VINet: Visual-Inertial Odometry as a Sequence-to-Sequence Learning Problem
• SE3 : http://www.seas.upenn.edu/~meam620/slides/kinematicsI.pdf

Codes

• FlowNet

  • https://lmb.informatik.uni-freiburg.de/resources/software.php
  • https://github.com/jgorgenucsd/corr_tf
  • https://github.com/lmb-freiburg/flownet2
  • https://github.com/sampepose/flownet2-tf

• DeepVO

  • https://github.com/sladebot/deepvo
  • https://github.com/themightyoarfish/deepVO
  • https://github.com/fshamshirdar/DeepVO (pytorch)

• VINet

  • https://github.com/HTLife/VINet

• SE3 Layer Implementation Example : https://github.com/ankurhanda/tf-new-op-example

• GVNN : https://arxiv.org/pdf/1607.07405.pdf