search

JialeCao001 / D2Det

D2Det: Towards High Quality Object Detection and Instance Segmentation (CVPR2020)
https://openaccess.thecvf.com/content_CVPR_2020/papers/Cao_D2Det_Towards_High_Quality_Object_Detection_and_Instance_Segmentation_CVPR_2020_paper.pdf
MIT License
297 stars 86 forks source link

ImportError: made by deform_conv_cuda #39

Open Atom-Jinny opened 3 years ago

Atom-Jinny commented 3 years ago

I followed the tutorial to set the D2Det enviorment by the dockerfile and wanted to train custom dataset. the code and the steps are following: folder :D2Det docker build -t mmdetection docker/ make the container form image mmdetection ,and upload the data docker run --gpus all -it --name mm -v {DIR}/D2Det-master/data/:/mmdetection/data mmdetection check the packages version:pytorch 1.3.0 mmdet 2.10 python 3.6 image and docker don't have mmcv packages image mmcv version link image the mmdetection/mmdet not exist D2Det/mmdet/ops so try to copy D2D to docker docker cp D2Det-master mm:/ python setup.py develop and run tools/train.py image ImportError: cannot import name 'deform_conv_cuda'

the floder /mmdet/ops/dcn image and floder /mmdet/ops/dcn/src image

what happen about it ?how can I solve it ?maybe the problem about compile Hope your answers

JialeCao001 commented 3 years ago

@Atom-Jinny I am not sure that it supports docker.

pratikadarsh commented 3 years ago

maybe related issue #34 .