def transform_XXXTrainData(z):
z.include_files.append( "arrayobject.h" ) # to get NumPy's flags
z.include_files.append( "ndarray.hpp" )
z.mem_fun('get_vectors')._transformer_creators.extend([
FT.input_array1d('values'), FT.input_array1d('missing'), FT.input_array1d('responses')])
for t in (
'get_cv_labels', 'get_sample_indices', 'get_cat_var_data', 'get_ord_var_data',
):
z.mem_funs(t).exclude()
z.add_declaration_code('''
static bp::object CLASS_NAME_get_cat_var_data(bp::object const &bpinst, CvDTreeNode* n, int vi)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_cat_var_data(n, vi, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::object CLASS_NAME_get_cv_labels(bp::object const &bpinst, CvDTreeNode* n)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_cv_labels(n, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::object CLASS_NAME_get_sample_indices(bp::object const &bpinst, CvDTreeNode* n)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_sample_indices(n, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::tuple CLASS_NAME_get_ord_var_data(bp::object const &bpinst, CvDTreeNode* n, int vi)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
std::vector<int> sample_indices(size);
sdcpp::ndarray result1 = sdcpp::simplenew_ndarray(1, &size, NPY_FLOAT);
sdcpp::ndarray result2 = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
float *in_arr1 = (float*)result1.data();
int *in_arr2 = (int*)result2.data();
const float *out_arr1;
const int *out_arr2;
inst.get_ord_var_data(n, vi, in_arr1, in_arr2, &out_arr1, &out_arr2, &sample_indices[0]);
bp::object obj1, obj2;
if(out_arr1 && out_arr1 != in_arr1)
{
result1 = sdcpp::new_ndarray1d(size, NPY_FLOAT, (void *)out_arr1);
bp::objects::make_nurse_and_patient(result1.get_obj().ptr(), bpinst.ptr());
}
if(out_arr1) obj1 = result1.get_obj();
if(out_arr2 && out_arr2 != in_arr2)
{
result2 = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr2);
bp::objects::make_nurse_and_patient(result2.get_obj().ptr(), bpinst.ptr());
}
if(out_arr2) obj2 = result2.get_obj();
return bp::make_tuple(obj1, obj2);
}
'''.replace("CLASS_NAME", z.alias).replace("CLASS_TYPE", z.pds))
z.add_registration_code('def("get_cat_var_data", &::CLASS_NAME_get_cat_var_data, (bp::arg("self"), bp::arg("n"), bp::arg("vi")))'.replace("CLASS_NAME", z.alias))
z.add_registration_code('def("get_cv_labels", &::CLASS_NAME_get_cv_labels, (bp::arg("self"), bp::arg("n")))'.replace("CLASS_NAME", z.alias))
z.add_registration_code('def("get_sample_indices", &::CLASS_NAME_get_sample_indices, (bp::arg("self"), bp::arg("n")))'.replace("CLASS_NAME", z.alias))
z.add_registration_code('def("get_ord_var_data", &::CLASS_NAME_get_ord_var_data, (bp::arg("self"), bp::arg("n"), bp::arg("vi")))'.replace("CLASS_NAME", z.alias))
''')
for z in (
'cvCalcOpticalFlowLK', 'cvCalcOpticalFlowBM', 'cvCalcOpticalFlowHS',
'cvEstimateRigidTransform', 'cvCalcOpticalFlowFarneback',
'cvUpdateMotionHistory', 'cvCalcMotionGradient', 'cvCalcGlobalOrientation',
'cvAcc', 'cvSquareAcc', 'cvMultiplyAcc', 'cvRunningAvg',
):
sb.mb.free_fun(z).include()
# cvCalcOpticalFlowPyrLK
FT.expose_func(sb.mb.free_fun('cvCalcOpticalFlowPyrLK'), return_pointee=False, transformer_creators=[
FT.input_array1d('prev_features', 'count', output_arrays=[('curr_features','1'),
('status','1'), ('track_error','1')])])
# cvCalcAffineFlowPyrLK
FT.expose_func(sb.mb.free_fun('cvCalcAffineFlowPyrLK'), return_pointee=False, transformer_creators=[
FT.input_array1d('prev_features', 'count', output_arrays=[('curr_features','1'),
('matrices','1'), ('status','1'), ('track_error','1')])])
# cvSegmentMotion
FT.expose_func(sb.mb.free_fun('cvSegmentMotion'), ward_indices=(3,))
# Planar Subdivisions
sb.cc.write('''
#-----------------------------------------------------------------------------
# Planar Subdivisions
#-----------------------------------------------------------------------------
sb.finalize_class(z)
# CvDTreeNode
z = sb.mb.class_('CvDTreeNode')
sb.init_class(z)
for t in ('parent', 'left', 'right', 'split'):
FT.expose_member_as_pointee(z, t)
for t in ('num_valid', 'cv_Tn', 'cv_node_risk', 'cv_node_error'):
z.var(t).exclude() # members for internal use, no need to expose
sb.finalize_class(z)
# CvDTreeParams
z = sb.mb.class_('CvDTreeParams')
sb.init_class(z)
FT.expose_member_as_ndarray1d(z, 'priors', 'inst.max_categories')
z.constructor(lambda x: len(x.arguments) > 1)._transformer_creators.append(FT.input_array1d('_priors'))
sb.finalize_class(z)
# ...TrainData functions
def transform_XXXTrainData(z):
z.include_files.append( "arrayobject.h" ) # to get NumPy's flags
z.include_files.append( "ndarray.hpp" )
z.mem_fun('get_vectors')._transformer_creators.extend([
FT.input_array1d('values'), FT.input_array1d('missing'), FT.input_array1d('responses')])
for t in (
'get_cv_labels', 'get_sample_indices', 'get_cat_var_data', 'get_ord_var_data',
):
z.mem_funs(t).exclude()
z.add_declaration_code('''
static bp::object CLASS_NAME_get_cat_var_data(bp::object const &bpinst, CvDTreeNode* n, int vi)
{
def beautify_func_list(func_list):
func_list = [f for f in func_list if not f.ignore]
# fix default values
# don't remove std::vector default values, old compilers _need_ std::allocator removed
for f in func_list:
for arg in f.arguments:
if isinstance(arg.default_value, str):
repl_list = {
'std::basic_string<char, std::char_traits<char>, std::allocator<char> >': 'std::string',
'std::vector<cv::Point_<int>, std::allocator<cv::Point_<int> > >': 'std::vector<cv::Point>',
'std::vector<cv::Scalar_<double>, std::allocator<cv::Scalar_<double> > >': 'std::vector<cv::Scalar>',
'std::vector<int, std::allocator<int> >': 'std::vector<int>',
'std::vector<cv::Vec<int, 4>, std::allocator<cv::Vec<int, 4> > >': 'std::vector<cv::Vec4i>',
}
for z in repl_list:
arg.default_value = arg.default_value.replace(z, repl_list[z])
# one-to-one function argument
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
pds = _c.unique_pds(arg.type.partial_decl_string)
if pds in _c.c2cpp:
f._transformer_creators.append(_FT.input_as_FixType(pds, _c.c2cpp[pds], arg.name))
elif pds in ['CvRNG *', 'CvRNG &', 'CvRNG cosnt *', 'CvRNG const &']:
f._transformer_creators.append(_FT.input_asRNG(arg.name))
elif pds in ['CvFileStorage *', 'CvFileStorage const *']:
f._transformer_creators.append(_FT.input_as_FileStorage(arg.name))
elif pds in ['CvFileNode *', 'CvFileNode const *']:
f._transformer_creators.append(_FT.input_as_FileNode(arg.name))
elif pds in ['CvMemStorage *', 'CvMemStorage const *']:
f._transformer_creators.append(_FT.input_as_MemStorage(arg.name))
elif pds in ['CvSparseMat *', 'CvSparseMat &', 'CvSparseMat const *', 'CvSparseMat const &']:
f._transformer_creators.append(_FT.input_asSparseMat(arg.name))
elif pds in ["IplImage *", "IplImage const *", "CvArr *", "CvArr const *",
"CvMat *", "CvMat const *", "cv::Range const *"]:
f._transformer_creators.append(_FT.input_as_Mat(arg.name))
# function argument int *sizes and int dims
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'sizes' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == 'dims' and _D.is_integral(arg2.type):
f._transformer_creators.append(_FT.input_array1d('sizes', 'dims'))
break
if arg.name == '_sizes' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == '_ndims' and _D.is_integral(arg2.type):
f._transformer_creators.append(_FT.input_array1d('_sizes', '_ndims'))
break
if arg2.name == 'dims' and _D.is_integral(arg2.type):
f._transformer_creators.append(_FT.input_array1d('_sizes', 'dims'))
break
if arg.name == '_newsz' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == '_newndims' and _D.is_integral(arg2.type):
f._transformer_creators.append(_FT.input_array1d('_newsz', '_newndims'))
break
# function argument const CvPoint2D32f * src and const CvPoint2D32f * dst
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'src' and _D.is_pointer(arg.type) and 'CvPoint2D32f' in arg.type.decl_string:
for arg2 in f.arguments:
if arg2.name == 'dst' and _D.is_pointer(arg2.type) and 'CvPoint2D32f' in arg2.type.decl_string:
f._transformer_creators.append(_FT.input_array1d('src'))
f._transformer_creators.append(_FT.input_array1d('dst'))
break
# argument 'void *data'
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'data' and _D.is_void_pointer(arg.type):
f._transformer_creators.append(_FT.input_string(arg.name))
self.add_doc(f.name, "'data' is represented by a string")
# final step: apply all the function transformations
for f in func_list:
if len(f._transformer_creators) > 0:
sort_transformers(f)
f.add_transformation(*f._transformer_creators, **f._transformer_kwds)
if 'unique_function_name' in f._transformer_kwds:
f.transformations[0].unique_name = f._transformer_kwds['unique_function_name']
else:
s = f.transformations[0].unique_name
repl_dict = {
'operator()': '__call__',
}
for t in repl_dict:
if t in s:
s = s.replace(t, repl_dict[t])
f.transformations[0].unique_name = s
f.transformations[0].alias = repl_dict[t]
break
_c.add_decl_desc(f)
const int *sz = inst.size();
if(!sz) return bp::object();
std::vector<int> result(inst.dims());
for(i = 0; i < inst.dims(); ++i) result[i] = sz[i];
return bp::object(result);
}
''')
z.add_registration_code(
'def("size", &::SparseMat_size, (bp::arg("i")=bp::object(-1)))')
z1 = z.mem_fun(
lambda x: x.name == "begin" and "const" not in x.partial_decl_string)
z1.include()
z1.rename("__iter__")
z.mem_fun(lambda x: x.name == 'hash' and 'int const *' in x.arguments[0].type.decl_string) \
._transformer_creators.append(FT.input_array1d('idx'))
for z2 in z.mem_funs('erase'):
z2._transformer_creators.append(FT.output_type1('hashval'))
if z2.arguments[0].name == 'idx':
z2._transformer_creators.append(FT.input_array1d('idx'))
# Hdr
z1 = z.class_('Hdr')
sb.init_class(z1)
z1.constructor(lambda x: len(x.arguments) > 1).exclude()
z1.add_declaration_code('''
static boost::shared_ptr<cv::SparseMat::Hdr> SparseMat_Hdr__init1__(std::vector<int> const &_sizes, int _type)
{
return boost::shared_ptr<cv::SparseMat::Hdr>(new cv::SparseMat::Hdr(_sizes.size(), &_sizes[0], _type));
}
''')
z.include()
if z.arguments[1].name == 'keypoints':
z._transformer_creators.append(FT.arg_output('keypoints'))
else:
z._transformer_creators.append(FT.output(z.arguments[1].name))
FT.doc_output(z, z.arguments[1])
t = D.remove_const(D.remove_reference(z.arguments[1].type))
name = 'read_' + C_to_Python_name_dict[t.partial_decl_string]
z._transformer_kwds['alias'] = name
# getPerspectiveTransform, getAffineTransform
for t in ('getPerspectiveTransform', 'getAffineTransform'):
FT.expose_func(sb.mb.free_fun(t),
return_pointee=False,
transformer_creators=[
FT.input_array1d('src'),
FT.input_array1d('dst')
])
# goodFeaturesToTrack
FT.expose_func(sb.mb.free_fun('goodFeaturesToTrack'),
return_pointee=False,
transformer_creators=[FT.arg_output('corners')])
# 'HoughCircles', 'HoughLines', 'HoughLinesP'
FT.expose_func(sb.mb.free_fun('HoughCircles'),
return_pointee=False,
transformer_creators=[FT.arg_output('circles')])
FT.expose_func(sb.mb.free_fun('HoughLines'),
return_pointee=False,
transformer_creators=[FT.arg_output('lines')])
if(i >= 0) return bp::object(inst.size(i));
const int *sz = inst.size();
if(!sz) return bp::object();
std::vector<int> result(inst.dims());
for(i = 0; i < inst.dims(); ++i) result[i] = sz[i];
return bp::object(result);
}
''')
z.add_registration_code('def("size", &::SparseMat_size, (bp::arg("i")=bp::object(-1)))')
z1 = z.mem_fun(lambda x: x.name=="begin" and "const" not in x.partial_decl_string)
z1.include()
z1.rename("__iter__")
z.mem_fun(lambda x: x.name == 'hash' and 'int const *' in x.arguments[0].type.decl_string) \
._transformer_creators.append(FT.input_array1d('idx'))
for z2 in z.mem_funs('erase'):
z2._transformer_creators.append(FT.output_type1('hashval'))
if z2.arguments[0].name == 'idx':
z2._transformer_creators.append(FT.input_array1d('idx'))
# Hdr
z1 = z.class_('Hdr')
sb.init_class(z1)
z1.constructor(lambda x: len(x.arguments) > 1).exclude()
z1.add_declaration_code('''
static boost::shared_ptr<cv::SparseMat::Hdr> SparseMat_Hdr__init1__(std::vector<int> const &_sizes, int _type)
{
return boost::shared_ptr<cv::SparseMat::Hdr>(new cv::SparseMat::Hdr(_sizes.size(), &_sizes[0], _type));
}
''')
for z in sb.mb.free_funs(lambda x: x.name=='read' and \
x.arguments[0].type.partial_decl_string.startswith('::cv::FileNode')):
z.include()
if z.arguments[1].name=='keypoints':
z._transformer_creators.append(FT.arg_output('keypoints'))
else:
z._transformer_creators.append(FT.output(z.arguments[1].name))
FT.doc_output(z, z.arguments[1])
t = D.remove_const(D.remove_reference(z.arguments[1].type))
name = 'read_'+C_to_Python_name_dict[t.partial_decl_string]
z._transformer_kwds['alias'] = name
# getPerspectiveTransform, getAffineTransform
for t in ('getPerspectiveTransform', 'getAffineTransform'):
FT.expose_func(sb.mb.free_fun(t), return_pointee=False,
transformer_creators=[FT.input_array1d('src'), FT.input_array1d('dst')])
# goodFeaturesToTrack
FT.expose_func(sb.mb.free_fun('goodFeaturesToTrack'), return_pointee=False,
transformer_creators=[FT.arg_output('corners')])
# 'HoughCircles', 'HoughLines', 'HoughLinesP'
FT.expose_func(sb.mb.free_fun('HoughCircles'), return_pointee=False,
transformer_creators=[FT.arg_output('circles')])
FT.expose_func(sb.mb.free_fun('HoughLines'), return_pointee=False,
transformer_creators=[FT.arg_output('lines')])
FT.expose_func(sb.mb.free_fun('HoughLinesP'), return_pointee=False,
transformer_creators=[FT.arg_output('lines')])
# getOptimalNewCameraMatrix
FT.expose_func(sb.mb.free_fun('getOptimalNewCameraMatrix'), return_pointee=False,
def beautify_func_list(func_list):
func_list = [f for f in func_list if not f.ignore]
# fix default values
# don't remove std::vector default values, old compilers _need_ std::allocator removed
for f in func_list:
for arg in f.arguments:
if isinstance(arg.default_value, str):
repl_list = {
'std::basic_string<char, std::char_traits<char>, std::allocator<char> >':
'std::string',
'std::vector<cv::Point_<int>, std::allocator<cv::Point_<int> > >':
'std::vector<cv::Point>',
'std::vector<cv::Scalar_<double>, std::allocator<cv::Scalar_<double> > >':
'std::vector<cv::Scalar>',
'std::vector<int, std::allocator<int> >':
'std::vector<int>',
'std::vector<cv::Vec<int, 4>, std::allocator<cv::Vec<int, 4> > >':
'std::vector<cv::Vec4i>',
}
for z in repl_list:
arg.default_value = arg.default_value.replace(
z, repl_list[z])
# one-to-one function argument
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
pds = _c.unique_pds(arg.type.partial_decl_string)
if pds in _c.c2cpp:
f._transformer_creators.append(
_FT.input_as_FixType(pds, _c.c2cpp[pds], arg.name))
elif pds in [
'CvRNG *', 'CvRNG &', 'CvRNG cosnt *', 'CvRNG const &'
]:
f._transformer_creators.append(_FT.input_asRNG(arg.name))
elif pds in ['CvFileStorage *', 'CvFileStorage const *']:
f._transformer_creators.append(
_FT.input_as_FileStorage(arg.name))
elif pds in ['CvFileNode *', 'CvFileNode const *']:
f._transformer_creators.append(_FT.input_as_FileNode(arg.name))
elif pds in ['CvMemStorage *', 'CvMemStorage const *']:
f._transformer_creators.append(
_FT.input_as_MemStorage(arg.name))
elif pds in [
'CvSparseMat *', 'CvSparseMat &', 'CvSparseMat const *',
'CvSparseMat const &'
]:
f._transformer_creators.append(_FT.input_asSparseMat(arg.name))
elif pds in [
"IplImage *", "IplImage const *", "CvArr *",
"CvArr const *", "CvMat *", "CvMat const *",
"cv::Range const *"
]:
f._transformer_creators.append(_FT.input_as_Mat(arg.name))
# function argument int *sizes and int dims
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'sizes' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == 'dims' and _D.is_integral(arg2.type):
f._transformer_creators.append(
_FT.input_array1d('sizes', 'dims'))
break
if arg.name == '_sizes' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == '_ndims' and _D.is_integral(arg2.type):
f._transformer_creators.append(
_FT.input_array1d('_sizes', '_ndims'))
break
if arg2.name == 'dims' and _D.is_integral(arg2.type):
f._transformer_creators.append(
_FT.input_array1d('_sizes', 'dims'))
break
if arg.name == '_newsz' and _D.is_pointer(arg.type):
for arg2 in f.arguments:
if arg2.name == '_newndims' and _D.is_integral(arg2.type):
f._transformer_creators.append(
_FT.input_array1d('_newsz', '_newndims'))
break
# function argument const CvPoint2D32f * src and const CvPoint2D32f * dst
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'src' and _D.is_pointer(
arg.type) and 'CvPoint2D32f' in arg.type.decl_string:
for arg2 in f.arguments:
if arg2.name == 'dst' and _D.is_pointer(
arg2.type
) and 'CvPoint2D32f' in arg2.type.decl_string:
f._transformer_creators.append(
_FT.input_array1d('src'))
f._transformer_creators.append(
_FT.input_array1d('dst'))
break
# argument 'void *data'
for f in func_list:
for arg in f.arguments:
if is_arg_touched(f, arg.name):
continue
if arg.name == 'data' and _D.is_void_pointer(arg.type):
f._transformer_creators.append(_FT.input_string(arg.name))
self.add_doc(f.name, "'data' is represented by a string")
# final step: apply all the function transformations
for f in func_list:
if len(f._transformer_creators) > 0:
sort_transformers(f)
f.add_transformation(*f._transformer_creators,
**f._transformer_kwds)
if 'unique_function_name' in f._transformer_kwds:
f.transformations[0].unique_name = f._transformer_kwds[
'unique_function_name']
else:
s = f.transformations[0].unique_name
repl_dict = {
'operator()': '__call__',
}
for t in repl_dict:
if t in s:
s = s.replace(t, repl_dict[t])
f.transformations[0].unique_name = s
f.transformations[0].alias = repl_dict[t]
break
_c.add_decl_desc(f)
def transform_XXXTrainData(z):
z.include_files.append("arrayobject.h") # to get NumPy's flags
z.include_files.append("ndarray.hpp")
z.mem_fun('get_vectors')._transformer_creators.extend([
FT.input_array1d('values'),
FT.input_array1d('missing'),
FT.input_array1d('responses')
])
for t in (
'get_cv_labels',
'get_sample_indices',
'get_cat_var_data',
'get_ord_var_data',
):
z.mem_funs(t).exclude()
z.add_declaration_code('''
static bp::object CLASS_NAME_get_cat_var_data(bp::object const &bpinst, CvDTreeNode* n, int vi)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_cat_var_data(n, vi, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::object CLASS_NAME_get_cv_labels(bp::object const &bpinst, CvDTreeNode* n)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_cv_labels(n, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::object CLASS_NAME_get_sample_indices(bp::object const &bpinst, CvDTreeNode* n)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
sdcpp::ndarray result = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
int *in_arr = (int*)result.data();
const int *out_arr = inst.get_sample_indices(n, in_arr);
if(!out_arr) return bp::object();
if(out_arr == in_arr) return result.get_obj();
result = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr);
bp::objects::make_nurse_and_patient(result.get_obj().ptr(), bpinst.ptr());
return result.get_obj();
}
static bp::tuple CLASS_NAME_get_ord_var_data(bp::object const &bpinst, CvDTreeNode* n, int vi)
{
CLASS_TYPE &inst = bp::extract<CLASS_TYPE &>(bpinst);
int size = n->sample_count;
std::vector<int> sample_indices(size);
sdcpp::ndarray result1 = sdcpp::simplenew_ndarray(1, &size, NPY_FLOAT);
sdcpp::ndarray result2 = sdcpp::simplenew_ndarray(1, &size, NPY_INT);
float *in_arr1 = (float*)result1.data();
int *in_arr2 = (int*)result2.data();
const float *out_arr1;
const int *out_arr2;
inst.get_ord_var_data(n, vi, in_arr1, in_arr2, &out_arr1, &out_arr2, &sample_indices[0]);
bp::object obj1, obj2;
if(out_arr1 && out_arr1 != in_arr1)
{
result1 = sdcpp::new_ndarray1d(size, NPY_FLOAT, (void *)out_arr1);
bp::objects::make_nurse_and_patient(result1.get_obj().ptr(), bpinst.ptr());
}
if(out_arr1) obj1 = result1.get_obj();
if(out_arr2 && out_arr2 != in_arr2)
{
result2 = sdcpp::new_ndarray1d(size, NPY_INT, (void *)out_arr2);
bp::objects::make_nurse_and_patient(result2.get_obj().ptr(), bpinst.ptr());
}
if(out_arr2) obj2 = result2.get_obj();
return bp::make_tuple(obj1, obj2);
}
'''.replace("CLASS_NAME", z.alias).replace("CLASS_TYPE", z.pds))
z.add_registration_code(
'def("get_cat_var_data", &::CLASS_NAME_get_cat_var_data, (bp::arg("self"), bp::arg("n"), bp::arg("vi")))'
.replace("CLASS_NAME", z.alias))
z.add_registration_code(
'def("get_cv_labels", &::CLASS_NAME_get_cv_labels, (bp::arg("self"), bp::arg("n")))'
.replace("CLASS_NAME", z.alias))
z.add_registration_code(
'def("get_sample_indices", &::CLASS_NAME_get_sample_indices, (bp::arg("self"), bp::arg("n")))'
.replace("CLASS_NAME", z.alias))
z.add_registration_code(
'def("get_ord_var_data", &::CLASS_NAME_get_ord_var_data, (bp::arg("self"), bp::arg("n"), bp::arg("vi")))'
.replace("CLASS_NAME", z.alias))
# CvDTreeNode
z = sb.mb.class_('CvDTreeNode')
sb.init_class(z)
for t in ('parent', 'left', 'right', 'split'):
FT.expose_member_as_pointee(z, t)
for t in ('num_valid', 'cv_Tn', 'cv_node_risk', 'cv_node_error'):
z.var(t).exclude() # members for internal use, no need to expose
sb.finalize_class(z)
# CvDTreeParams
z = sb.mb.class_('CvDTreeParams')
sb.init_class(z)
FT.expose_member_as_ndarray1d(z, 'priors', 'inst.max_categories')
z.constructor(lambda x: len(x.arguments) > 1)._transformer_creators.append(
FT.input_array1d('_priors'))
sb.finalize_class(z)
# ...TrainData functions
def transform_XXXTrainData(z):
z.include_files.append("arrayobject.h") # to get NumPy's flags
z.include_files.append("ndarray.hpp")
z.mem_fun('get_vectors')._transformer_creators.extend([
FT.input_array1d('values'),
FT.input_array1d('missing'),
FT.input_array1d('responses')
])
for t in (
'get_cv_labels',
'get_sample_indices',
'cvCalcMotionGradient',
'cvCalcGlobalOrientation',
'cvAcc',
'cvSquareAcc',
'cvMultiplyAcc',
'cvRunningAvg',
):
sb.mb.free_fun(z).include()
# cvCalcOpticalFlowPyrLK
FT.expose_func(sb.mb.free_fun('cvCalcOpticalFlowPyrLK'),
return_pointee=False,
transformer_creators=[
FT.input_array1d('prev_features',
'count',
output_arrays=[('curr_features', '1'),
('status', '1'),
('track_error', '1')])
])
# cvCalcAffineFlowPyrLK
FT.expose_func(sb.mb.free_fun('cvCalcAffineFlowPyrLK'),
return_pointee=False,
transformer_creators=[
FT.input_array1d('prev_features',
'count',
output_arrays=[('curr_features', '1'),
('matrices', '1'),
('status', '1'),
('track_error', '1')])
])