I'm trying to convert a caffe model to Core ML but I'm getting this error:
Message type "caffe.LayerParameter" has no field named "normalize_bbox_param".
What should I do?
Thanks for trying out the Beta! Models trained using standard Caffe installation will convert with Core ML converters, but from the logs, it looks like you might be using a different fork of Caffe. “normalize_bbox_param” or “norm_param” is a parameter belonging to a layer called “NormalizeBBox". This version of caffe seems to have come from here: https://github.com/opencv/opencv_contrib/blob/master/modules/dnn/src/caffe/caffe.proto, which is different from the BVLC Caffe.
Core ML supports the official version of caffe: http://caffe.berkeleyvision.org and the list of layers in BVLC Caffe are available here: https://github.com/BVLC/caffe/blob/master/src/caffe/proto/caffe.proto
Note: It looks like the parameters “normalize_bbox_param” and “norm_param” are the same parameters with different names on different forks of Caffe: https://github.com/opencv/opencv_contrib/issues/935