Caffe - Use ".caffemodel " file with Python->C++ file translated layers - caffe

Question

I have a .prototxt file and a .caffemodel file, for a network I built and trained in py-caffe, using self developed Python layers.
Later, I implemented the same layers for the C++ version of Caffe.
The question is whether it is possible to only change the layer types (from the Python ones to the C++ ones) in the .prototxt file, while using the same already trained .caffemodel file, in order to load and use the trained network, this time with the C++ layers?

Answer 1

Yes, of course it's possible. In case you have layer parameters, you should add a new layer_param message to caffe.proto, and read them accordingly in your cpp layer.
For example, PVANet caffe repository has implementation for the original Python proposal layer.
As you can see in caffe.proto, a new message was added to hold the Python layer's parameters -
// Message that stores parameters used by ProposalLayer
message ProposalParameter {
optional uint32 feat_stride = 1 [default = 16];
optional uint32 base_size = 2 [default = 16];
optional uint32 min_size = 3 [default = 16];
repeated float ratio = 4;
repeated float scale = 5;
optional uint32 pre_nms_topn = 6 [default = 6000];
optional uint32 post_nms_topn = 7 [default = 300];
optional float nms_thresh = 8 [default = 0.7];
}
The original train / test prototxt contains the following layer:
layer {
name: 'proposal'
type: 'Python'
bottom: 'rpn_cls_prob_reshape'
bottom: 'rpn_bbox_pred'
bottom: 'im_info'
top: 'rpn_rois'
top: 'rpn_scores'
include { phase: TRAIN }
python_param {
module: 'rpn.proposal_layer'
layer: 'ProposalLayer'
param_str: "{'feat_stride': 16, 'ratios': [0.333, 0.5, 0.667, 1, 1.5, 2, 3], 'scales': [2, 3, 5, 9, 16, 32]}"
}
}
When the new cpp one looks like:
layer {
name: 'proposal'
type: 'proposal'
bottom: 'rpn_cls_prob_reshape'
bottom: 'rpn_bbox_pred'
bottom: 'im_info'
top: 'rpn_rois'
top: 'rpn_scores'
include { phase: TRAIN }
proposal_param {
feat_stride: 16
ratios: 0.333, 0.5, 0.667, 1, 1.5, 2, 3
scales: 2, 3, 5, 9, 16, 32
}
}
etc.