Yi Li
Yi Li
Maybe you can try printing the path before you read the file and see whether the one leads to this error have some problem. It seems that the program fails...
Hi, @gaopeng-eugene , were you running it in debug mode?
Hi, have you > Copy operators in ./fcis/operator_cxx to $(YOUR_MXNET_FOLDER)/src/operator/contrib and recompile MXNet.
Hi, @phexic , have you updated your mxnet? [mxnet_op.h](https://github.com/dmlc/mxnet/blob/master/src/operator/mxnet_op.h)
the training images are read through this function https://github.com/msracver/FCIS/blob/master/lib/segmentation/segmentation.py#L20
@hixuehai You can try to modify https://github.com/msracver/FCIS/blob/master/lib/utils/image.py#L25 or https://github.com/msracver/FCIS/blob/master/lib/utils/image.py#L56 to meet your needs.
@xingbowei You can refer to the config and symbol files used for coco.
@xingbowei you can refer to code from L174 to L180, https://github.com/msracver/FCIS/blob/master/lib/dataset/coco.py#L174
Because we need multi-gpu support with python api, which caffe doesn't provide, at least at that time.
Hi, @fangxu622 did you do the following steps before building mxnet? ``` git checkout 62ecb60 git submodule update ```