No "reshape_as_input" for class "caffe.Net"

[AlexofNTU]

Dear all:
I am trying to use Faster R-CNN

However, there is an error message coming from executing script_faster_rcnn_demo.m
The error message is
No appropriate method, property, or field 'reshape_as_input' for class 'caffe.Net'
in proposal_im_detect.m

Can anyone give me a clue about how to solve this issue?
Thanks

Alex

[choasup]

You should amend the 'detection_test.prototxt' like the 'detection_test.prototxt' in output that you download.

[wangxiao5791509]

@AlexofNTU I met the same problem as yours, and I have another question, i.e.
It stucked in Stage 1, and the terminal shown me this:
File not found: ./models/rpn_prototxts/ZF/train_vali.prototxt
But the file really existing in this path, it really confused me. Waiting for answers online ...

Thanks for every kind answers ...

[Zy-Zhang]

@wangxiao5791509 I've met this problem before. Just try to rewrite all '' into '//' in solver.prototxt file. I fixed it by this method.

[Mato98]

Did anyone solve the problem "No appropriate method, property, or field 'reshape_as_input' for class 'caffe.Net'" in proposal_im_detect.m? I get the same when running "script_faster_rcnn_demo.m"

What is meant by

amend the 'detection_test.prototxt' like the 'detection_test.prototxt' in output that you download

i only have one "detection_test.ptototxt" file and it's the one i downloaded with the matlab command "fetch_faster_rcnn_final_model"

Thanks
Tobi

[Thanhpqt92]

I can't fix this problem by the solution of @Izzy88.

Please help me to solve this one.

[sizhangyu]

I met the same problem. Has anyone solved it?

[wenguanwang]

Hi, everyone,

I suggest you should use the modified caffe offered by fasterrcnn of Ren, instead of the original caffe. The function 'net.reshape' is ONLY offered in the caffe of faster rcnn.

[wenguanwang]

Hi, guys,

I have checked the caffe code of faster rcnn. That is totally due to the version of caffe. As I mentioned, you should:

1. use the caffe of faster rcnn and recomplie;

Or

2. modify caffe/matlab/+caffe/Net.m by adding this function: (just copy following codes into the end of Net.m)

function reshape_as_input(self, input_data)
CHECK(iscell(input_data), 'input_data must be a cell array');
CHECK(length(input_data) == length(self.inputs), ...
'input data cell length must match input blob number');
% reshape input blobs
for n = 1:length(self.inputs)
if isempty(input_data{n})
continue;
end
input_data_size = size(input_data{n});
input_data_size_extended = [input_data_size, ones(1, 4 - length(input_data_size))];
self.blobs(self.inputs{n}).reshape(input_data_size_extended);
end
self.reshape();
end