I fine tuned the VGG16 network using KITTI data and once the solver finished the optimization, the caffemodel and solver state were saved as usual. However, when I try to use the caffemodel to extract features, I get an error.
E1212 13:49:03.041193 18874 extract_features.cpp:62] Using CPU
F1212 13:49:07.289798 18874 upgrade_proto.cpp:95] Check failed: ReadProtoFromBinaryFile(param_file, param) Failed to parse NetParameter file: kittitrained.caffemodel
*** Check failure stack trace: ***
@ 0x7ffff04cec8d google::LogMessage::Fail()
@ 0x7ffff04d3bba google::LogMessage::SendToLog()
@ 0x7ffff04d1021 google::LogMessage::Flush()
@ 0x7ffff04d13dd google::LogMessageFatal::~LogMessageFatal()
@ 0x7ffff745d878 caffe::ReadNetParamsFromBinaryFileOrDie()
@ 0x7ffff7425a3f caffe::Net<>::CopyTrainedLayersFromBinaryProto()
@ 0x7ffff7425b16 caffe::Net<>::CopyTrainedLayersFrom()
@ 0x4098c0 feature_extraction_pipeline<>()
@ 0x32de01ed5d (unknown)
@ 0x4055b9 (unknown)
Aborted