Speed OpenFaceOffline - Githubissues

TadasBaltrusaitis / OpenFace

OpenFace – a state-of-the art tool intended for facial landmark detection, head pose estimation, facial action unit recognition, and eye-gaze estimation.

Other

6.72k stars 1.83k forks source link

Speed OpenFaceOffline #854

Open sergio-dl opened 4 years ago

sergio-dl commented 4 years ago

Hello,

I am very new in using OpenFace... When I use OpenfaceOffline, I get only a video and a CSV file separately. Is it possible to record all the information that openfaceoffline shows when it is processing ? I tried to capture the screen but the speed of the video is much longer than the original video. Thus, it is impossible to synchronies with the original video. On other words, I need to have a video similar to what openfaceoffline shows when processing but at the same speed than the original video. How could I do that ? Thanks for your help Regards Sergio

TadasBaltrusaitis commented 4 years ago

Hi,

Is there a particular piece of information you believe is missing? All the information should be there, just in a different format. if it is missing then that would be a bug.

OpenFaceOffline outputs tracked video/image + CSV + aligned faces + HOG features.

Thanks, Tadas

amhelmi commented 4 years ago

Hi Tadas,

Thanks for your awesome work so far. I'm not sure if the above poster is asking the same question I have but it seems related.

The csv and video files are continually being written in OpenfaceOffline but cannot be accessed until it stops recording. Is there a way to access that information or the video for doing further processing in real time?

Cheers, Ameer

TadasBaltrusaitis commented 4 years ago

There are several ways to access that information live:

Write your code that needs to access it directly in the .cpp files (e.g. using FaceLandmarkVid.cpp or FeatureExtraction.cpp as starting points)
Look at how HeadPoseLive streams head pose and eye gaze information that can then be read by a python scripts using ZeroMQ (this is currently only supporting head pose and eye gaze, but could be extended to other signals with a bit of code)

ashoorie commented 3 years ago

Just a quick note: I ran testing_head_pose.py to access the live information from HeasPoseLive and got the following two errors:

-The first error was: TypeError: unicode not allowed, use setsockopt_string As the error message suggests, to resolve this error you could modify line 15: socket.setsockopt_string(zmq.SUBSCRIBE, topic_filter) Alternatively, you could modify line 14: topic_filter = b"HeadPose:"

-The second error message was: TypeError: a bytes-like object is required, not 'str' It could be resolved by inserting letter b in line 19: head_pose = head_pose[9:].split(b',')

I hope this helps.

johnbone25 commented 3 years ago

Is it possible to invoke the head position or eye tracking in Python? And how to do? Thanks