search

DenisTome / Lifting-from-the-Deep-release

Implementation of "Lifting from the Deep: Convolutional 3D Pose Estimation from a Single Image"
https://denistome.github.io/papers/lifting-from-the-deep
GNU General Public License v3.0
450 stars 132 forks source link

Error when running scripts on cpu #17

Closed yashiro32 closed 6 years ago

yashiro32 commented 6 years ago

I got this error: "Segmentation fault (core dumped)" when running your scripts on the cpu.

GajjarMihir commented 6 years ago

Did you first run the setup.sh file?

yashiro32 commented 6 years ago

Yes, I ran the setup.sh to download the trained models and to install the external utilities.

The scripts works on tensorflow-gpu but not on tensorflow-cpu and gotten the error I stated previously.

GajjarMihir commented 6 years ago

Ohh well..!! I ran it on tensorflow-gpu. So I guess I didnt get any error. The script works perfectly well but takes more time. I am trying to optimize it.

yashiro32 commented 6 years ago

It's on tensorflow-cpu that is getting the error. I ran it on cpu because the scripts need 4.9 GB of memory to run and my the memory on my gpu is insufficient.

DenisTome commented 6 years ago

I've uploaded a faster version that should also require less memory in the GPU. If that it's still not enough, you can try to reduce the batch size in the file config.py which would reduce further the space needed to allocate the graph.

p.s. The network is not very fast if the code is run on a single image. That it's tf problem that has to initialize some stuff the first time you call the pose_estimator.estimate(image) method. If this is executed on other images then it runs at 0.34 secs per frame on a Titan X.

MritulaC commented 5 years ago

setup.sh can be run from git bash on windows.

or

Download the *.tar.tz files manually by pasting the URLs from the setup.sh files and excecute each instruction line by line in git bash.

This is how I overcame the same issue