咨询faster_rcnn转ncnn
Opened this issue · 8 comments
你好，我通过ONNX转的NCNN，会出现很多不支持的层，想问一下您是怎么转换的，谢谢；
largest=1
sorted=1
Cast not supported yet!
to=7
Range not supported yet!
Shape not supported yet!
Expand not supported yet!
Shape not supported yet!
Cast not supported yet!
to=7
Gather not supported yet!
axis=0
Unsupported unsqueeze axes !
Unknown data type 0
Cast not supported yet!
to=7
Gather not supported yet!
axis=0
Unsupported unsqueeze axes !
Unknown data type 0
Cast not supported yet!
to=7
Gather not supported yet!
axis=0
Unsupported unsqueeze axes !
Unknown data type 0
Cast not supported yet!
Converting ONNX can cause problems. I have often come across them.
I have used the hints given by nihui in it's code.
In case of ZF_fatser_rcnn_final, you can download the weights by the link given in the README.md.
The param file used is found at the repo (ZF_faster_rcnn_final.proto).
Hello @Qengineering @zishui-wu
I'm facing error while converting to ncnn. Is there any tip from your side where I'm making mistake? Thank you for your answer
Unsupported split axis !
Unsupported squeeze axes !
Shape not supported yet!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Unsupported split axis !
Unsupported squeeze axes !
Shape not supported yet!
Cast not supported yet!
# to=1
Cast not supported yet!
# to=1
Unsupported unsqueeze axes !
Shape not supported yet!
Gather not supported yet!
# axis=0
Cast not supported yet!
# to=1
Cast not supported yet!
# to=1
Cast not supported yet!
# to=1
Shape not supported yet!
Gather not supported yet!
# axis=0
Cast not supported yet!
# to=1
Cast not supported yet!
# to=1
Cast not supported yet!
# to=1
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Shape not supported yet!
Cast not supported yet!
# to=7
Unknown data type 0
Unsupported Resize scales and sizes are all empty!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Unsupported unsqueeze axes !
Cast not supported yet!
# to=1
Cast not supported yet!
# to=7
Shape not supported yet!
Gather not supported yet!
# axis=0
Unsupported unsqueeze axes !
Cast not supported yet!
# to=1
Cast not supported yet!
# to=7
Shape not supported yet!
Gather not supported yet!
# axis=0
Unsupported unsqueeze axes !
Cast not supported yet!
# to=1
Cast not supported yet!
# to=7
Cast not supported yet!
# to=1
Cast not supported yet!
# to=7
Cast not supported yet!
# to=1
Cast not supported yet!
# to=7
Shape not supported yet!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Shape not supported yet!
Gather not supported yet!
# axis=0
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Unsupported unsqueeze axes !
Shape not supported yet!
Gather not supported yet!
# axis=0
Cast not supported yet!
# to=7
ConstantOfShape not supported yet!
# value 4
Unsupported slice step !
Cast not supported yet!
# to=7
Unknown data type 0
Segmentation fault
I'm sorry to say, but you are hitting the famous Unsupported errors.
In other words, the current version of ncnn doesn't have an implementation for the mentioned operation.
It happens all time, since the deep learning community is evolving and new operations are developed.
Frameworks, like ncnn, need to maintain their code in order to keep up with the present state of art.
It's almost an impossible job.
thank you for your answer means it's not possible to run on NCNN or can I try with older versions ??
@zishui-wu
Using ncnn with the models you want to convert is not possible.
Older versions are not an option, because they have less functionality.
but then how did you convert and make possibly run on ncnn?
Just as I told @apanand14, I've used existing models.
In case of ZF_fatser_rcnn_final, you can download the weights by the link given in the README.md.
The param file used is found at the repo (ZF_faster_rcnn_final.proto).
I got it. Thank you for the response.