Skip to content

serge-m/deepflow

Repository files navigation

FastDeepFlow, v1.0.1

Our code is mentioned only for scientific or personal use.
Please contact us for commercial use.

### Compiling ###

Simply type 'make' and the code must compile.
The program was only tested under a 64-bit Linux distribution.
We do not give any support for other OS.
This version use SSE instructions and thus require a compatible CPU.
Please consider using DeepFlow instead of FastDeepFlow for others CPUs.

### Using FastDeepFlow ###

You can use FastDeepFlow without any input matches.
The general command to use FastDeepFlow is:
./fastdeepflow <im1> <im2> <filename.flo> [options]

FastDeepFlow takes benefit of SSE instructions to speed up the computation of DeepFlow.
The results are slightly different due to floating point approximations.
The endpoint difference is around 1e-4 pixels in average but can be around 1 pixel near discontinuities.

For a list of the available options, you can type ./fastdeepflow -h
You can basically set some parameters, or directly use the parameter we use in the ICCV 2013 paper for MPI-Sintel, Middlebury or KITTI.
The option -match and -matchf allows to give input matches, either from an input file or from stdin.

### Example including deep matching ###

To test FastDeepFlow including the deep matching, you need to download the deep matching code on Jerome Revaud's webpage: http://lear.inrialpes.fr/people/revaud
The matches output by deep matching can directly be piped to FastDeepFlow like in the following example:

<deepmatching_path>/deepmatching sintel1.png sintel2.png -iccv_settings | python <deepmatching_path>/rescore.py sintel1.png sintel2.png | ./fastdeepflow sintel1.png sintel2.png sintel.flo -matchf -sintel

It creates a .flo file in a usual format. 
For instance, code for reading and displaying such format is available with the Middlebury dataset.

The matches are similar to the one we use in ICCV'13 paper. If you need exactly the same, you can download them on the webpage.
If you want better results, just replace -iccv_settings by -improved_settings.

### Bugs and extensions ###

If you find bugs, etc., please feel free to contact me.
Contact details are available on my webpage.
http://lear.inrialpes.fr/people/pweinzae

### History ###

December 2013 v1.0.0
March 2014    v1.0.1: memory leak fix + support for png images
March 2014    FastDeepFlow_v1.0.1: faster (~2x) version using SSE instructions. Results are slightly different due to floating point approximations.

### LICENCE CONDITIONS ###

Copyright (C) 2013 Philippe Weinzaepfel

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <http://www.gnu.org/licenses/>.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages