jswulff / mrflow Goto Github PK
View Code? Open in Web Editor NEWCode for the paper "Optical Flow in Mostly Rigid Scenes" by Jonas Wulff, Laura Sevilla-Lara, Michael Black, CVPR 2017
License: Other
Code for the paper "Optical Flow in Mostly Rigid Scenes" by Jonas Wulff, Laura Sevilla-Lara, Michael Black, CVPR 2017
License: Other
Hi, I want to run mrflow.py for my frames. It is written that initialization with flows and rigidity image gives more accurate estimate. I want to ask if I give parameter --no_init, mrflow.py would still use CNN for estimating rigidity or not.
hello and thanks for the code!
are you planning to publish your code to generate the rigidity maps ?
or are there existing projects to do that already available .. ?
thanks in advance
luc
I got this error when trying to run the program on my own frames. I ran with the --no_init
option and provided three frames as specified. My images basically have a more or less still background with moving human subject (essentially the first few seconds of this video from Ellen Degeneres as a test for our later use cases)
==== EXCEPTION ====
Traceback (most recent call last):
File "mrflow.py", line 111, in compute_mrflow
params)
File "/Users/jg/repositories/mrflow/pipeline/compute_structure.py", line 429, in compute_structure
raise Exception('TooFewStructureMatches')
Exception: TooFewStructureMatches
I struggle to get some decent results on real life pictures/videos.
Blur looks like a significant issue bt maybe I a not using the algo properly. I am just usin the test syntax line from the readme so far.
Can you give some tweaks/advices to get better results if possible?
I attach the best test I got so far but I did expect some better rigidity map given the video you published a day ago.
I was following the instruction in the README.txt inside semantic_rigidity_cvpr2017. When I tried running python do_deeplab_segmentation.py TEST deeplab_deploy.prototxt deeplab_adam_0.0001_iter_1400.caffemodel .
it gives me the following error
FileNotFoundError: [Errno 2] No such file or directory: avg_train_image.npy
Can I use this code to calculate optical flow for a video recorded from a camera fixed on the dashboard of a car?
Also. can I modify the code to calculate the optical flow between 2 images only or this modification will affect the accuracy greatly?
Hi, first thanks for your work ad for sharing it.
I am trying to execute your example but I have the following error message. It seems it may relate to the env variable but I think I did export it properly. Can you help?below the shell outputs when executing:
tets:~/mrflow-master$ sudo python mrflow.py --no_init example/frame1.png example/frame2.png example/frame3.png
Traceback (most recent call last):
File "mrflow.py", line 11, in <module>
MRFLOW_HOME = os.environ['MRFLOW_HOME']
File "/usr/lib/python2.7/UserDict.py", line 40, in __getitem__
raise KeyError(key)
KeyError: 'MRFLOW_HOME'
tets:~/mrflow-master$ $MRFLOW_HOME/
bash: /home/tets/mrflow-master/: Is a directory
Tets
Thank you so much for sharing code.
When I run the demo code:
python mrflow.py --no_init example/frame1.png example/frame2.png example/frame3.png
It raised this error:
mrflow.py:189: RuntimeWarning: invalid value encountered in sqrt
flowdiff = np.sqrt((u-uinit)**2 + (v-vinit)**2)
sqrt
returns this error when we square root a negative number. (u-uinit)**2 + (v-vinit)**2
is always possitive, I don't know why it returns the error.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.