mx-maskrcnn
mx-maskrcnn copied to clipboard
TuSimple
An MXNet implementation of Mask R-CNN
When generating RPN detection, after training RPN1, the processing turned down. The error message is shown as below: ``` Traceback (most recent call last): File "train_alternate_mask_fpn.py", line 116, in main()...
Unable to download `bash scripts/download_res50.sh`？
After cp rcnn/CXX_OP/* incubator-mxnet/src/operator/ ,then compile,the errors occuring as follow. src/operator/./roi_align_v1-inl.h:336:9: error: ‘Fill’ was not declared in this scope Fill(s, outputs[0], kWriteTo, static_cast(0)); ^ /home/user/zhufan-mask/mx-maskrcnn/incubator-mxnet/mshadow/mshadow/./base.h:907:8: note: in definition of macro...
I follow the instruction: 1 cp rcnn/CXX_OP/* incubator-mxnet/src/operator/ 2 then cd incubator-mxnet 3 make but there is an error : "src/operator/./roi_align_v1-inl.h:342:7: error: ‘Fill’ was not declared in this scope"
how many steps in every epoch when train stage? And how many epoch when train rpn_stage1, fast_rcnn_stage1, rpn_stage2, faste_rcnn_stage2? Thanks!
I find that there maybe a bug in the ROIAlign operator. For example, x = mx.nd.array([[[[ 0., 1., 2., 3., 4., 5.], [ 6., 7., 8., 9., 10., 11.], [...
There is a bug in train_maskrcnn.py. # Line 74, in filter_roidb function valid = len(fg_inds) > 0 and len(bg_inds) > 0 should be: valid = len(fg_inds) > 0 or len(bg_inds)...
I find that there maybe a bug in the ROIAlign operator. For example, x = mx.nd.array([[[[ 0., 1., 2., 3., 4., 5.], [ 6., 7., 8., 9., 10., 11.], [...
I would like to use maskrcnn to do keypoint detection, what should i change? any suggestions are welcome.
