Error appears when running demo_rfcn.py
fighting-liu opened this issue · comments
I got the same problem.
are you using the coco branch?
I used the master branch, and caffe was built successfully. ( CUDA 8.0 with cuDNN ver5.1)
please read the readme
Thank you, the multi-gpus model is training now.
I have the same problem? Did you solve this problem, what's the method?
@zhanghaoinf @bharatsingh430
Hi, @liu09114 , I checked before when I used the detection model published by msra (models store in onedrive), the problem will occur. If you initialize and train your own model with resnet-101 (classification model pre-trained on ImageNet-1000), the problem will disappear. By the way, you need the coco branch to train your own model.
I think the sentence "If you want to use/train this model, please use the coco branch of this repository. " in the readme is important. :)
@liu09114, I guess there might be a slightly difference (maybe names, args ... ) in the detection layers between original py-rfcn and py-rfcn-multi-gpu version, which causes the failure of loading models trained with py-rfcn which I am not sure.
I have clone the coco branch by the following command
git clone --recursive https://github.com/bharatsingh430/py-R-FCN-multiGPU/ --branch coco --single-branch
But demo_rfcn.py
still not working. How to solve this issues @zhanghaoinf
I got this error too. so, what is the difference between py-rfcn and py-rfcn-multi-gpu version?
@xmubingo Hi, it has been half a year since I last use py-rfcn-mulit-gpu. I try to answer this as much as I can. After cloning this repo, switch to coco branch ("git branch -a" to list all branches, "git checkout -b experimental XXX", use "git branch" to view the change). Try coco model provided in this repo https://drive.google.com/file/d/0B6T5quL13CdHMGtMUWFFSXd2Ym8 instead of models downloaded from MSRA one drive.
@xmubingo
@shohan6
Note 2 differences:
-
The name of the model is different. In the code, name is
resnet101_rfcn_final
,actual iscoco_rfcn.caffemodel
-
Different paths, the code default is
cfg.models_dir = '/home/XXX/py-R-FCN-multiGPU-coco-branch/models/pascal_voc'
The actual is:
cfg.models_dir = '/home/XXX/py-R-FCN-multiGPU-coco-branch/models/coco'
Because this is the coco data set training model