FairMOT: On Windows 10 I get ModuleNotFoundError: No module named '_ext'

My lib versions

python 3.6.8
torch==1.4.0
torchsummary==1.5.1
torchvision==0.5.0

I got everything else installed but I get this error on Windows 10 when running demo.py

(pytorch) E:\Documents\Projects\FairMOT\src>python demo.py mot --load_model ../models/hrnetv2_w32_imagenet_pretrained.pth
Traceback (most recent call last):
  File "demo.py", line 14, in <module>
    from track import eval_seq
  File "E:\Documents\Projects\FairMOT\src\track.py", line 15, in <module>
    from tracker.multitracker import JDETracker
  File "E:\Documents\Projects\FairMOT\src\lib\tracker\multitracker.py", line 12, in <module>
    from lib.models.model import create_model, load_model
  File "E:\Documents\Projects\FairMOT\src\lib\models\model.py", line 11, in <module>
    from .networks.pose_dla_dcn import get_pose_net as get_dla_dcn
  File "E:\Documents\Projects\FairMOT\src\lib\models\networks\pose_dla_dcn.py", line 16, in <module>
    from .DCNv2.dcn_v2 import DCN
  File "E:\Documents\Projects\FairMOT\src\lib\models\networks\DCNv2\dcn_v2.py", line 13, in <module>
    import _ext as _backend
ModuleNotFoundError: No module named '_ext'

Though I have the .ext and .lib files copied in the same folder as the script

Also the python setup.py build develop didn’t succeed with the attached error out.txt

About this issue

Original URL
State: closed
Created 4 years ago
Comments: 70 (12 by maintainers)

Most upvoted comments

Finally I got the results

with the following command line

python demo.py mot --load_model ../models/all_hrnet_v2_w18.pth --input-video ../videos/mot16-03.mp4 --output-root ../results --arch hrnet_18 --reid_dim 128 --conf_thres 0.4

Thank you @ifzhang and @zubovskiii98 for helping out with this As for video output please let me know where to reach you @ifzhang my email is inbox at zubairahmed dot net closing issue

zubairahmed-ai on Apr 14, 2020

OK, I will upload some pretrained models of ResNet and HRNet and write some tutorials. When using ResNet and HRNet, you do not need to run build DCNv2.

ifzhang on Apr 11, 2020

@zubovskiii98 I am using CUDA 10

It was not very wise to discuss my issue in Windows 10 topic, because I am using Ubuntu 18

But in any case if you have followed the manual correctly, by doing:

pip install -r requirements.txt

Below fixes '_ext' error:
cd src/lib/models/networks/DCNv2
sh make.sh

And having right CUDA, everything should be fine, except…

I had an issue with mkdir_if_missing(), which is essentially os.makedirs(dir). It could not create a resulting directory, because of permissions error. Therefore, I have created the folders like videos, results, etc in root directory. (if not having ones) Then, I had to change following /src/track.py -> def main()

From:
 result_root = os.path.join(data_root, '..', 'results', exp_name)
To:
 result_root = os.path.join('..', 'results', exp_name)

Hope it helps

zubovskiii98 on Apr 14, 2020

hi, you can try python setup.py install develop

Pluto1314 on Apr 11, 2020