inline float gain(float value) const
{
if (value < 0.5f)
return bias(value * 2.0f) * 0.5f;
else
return bias(value * 2.0f - 1.0f) * 0.5f + 0.5f;
}
// apply a gain to x
float gain(float gain, float x)
{
if(x < 0.5f)
return bias(1.f-gain, 2.f*x)/2.f;
else
return 1.f - bias(1.f-gain, 2.f-2.f*x)/2.f;
}
void
Dynamic::AndBias::fromXml( QXmlStreamReader *reader )
{
while (!reader->atEnd()) {
reader->readNext();
if( reader->isStartElement() )
{
Dynamic::BiasPtr bias( Dynamic::BiasFactory::fromXml( reader ) );
if( bias )
{
appendBias( bias );
}
else
{
warning()<<"Unexpected xml start element"<<reader->name()<<"in input";
reader->skipCurrentElement();
}
}
else if( reader->isEndElement() )
{
break;
}
}
}
void TextureBuilder::createModules() {
QSharedPointer<ModuleDescriptor> mod;
foreach (mod,_modDesc) {
auto modPtr = mod.data();
if (this->useRandomFactors() && modPtr->enableRandom() ) {
modPtr->setBias(this->applyRandomFactor(modPtr->bias()) );
modPtr->setDispl(this->applyRandomFactor(modPtr->displ()) );
modPtr->setExp(this->applyRandomFactor(modPtr->exp()) );
modPtr->setFreq(this->applyRandomFactor(modPtr->freq()) );
modPtr->setLac(this->applyRandomFactor(modPtr->lac()) );
modPtr->setLbound(this->applyRandomFactor(modPtr->lBound()) );
modPtr->setPers(this->applyRandomFactor(modPtr->pers()) );
modPtr->setPow(this->applyRandomFactor(modPtr->pow()) );
modPtr->setRough(this->applyRandomFactor(modPtr->rough()) );
modPtr->setScale(this->applyRandomFactor(modPtr->scale()) );
modPtr->setValue(this->applyRandomFactor(modPtr->value()) );
modPtr->setX(this->applyRandomFactor(modPtr->x()) );
modPtr->setY(this->applyRandomFactor(modPtr->y()) );
modPtr->setZ(this->applyRandomFactor(modPtr->z()) );
}
mod.data()->setModules(_modules);
auto ptr = mod.data()->makeModule();
qDebug() << "Module " << modPtr->name() << " After module creation...";
modPtr->dumpModule();
_modules.insert(mod.data()->name(), ptr);
}
Dynamic::BiasPtr
Dynamic::AbstractBiasFactory::createFromXml( QXmlStreamReader *reader )
{
Dynamic::BiasPtr bias( createBias() );
bias->fromXml( reader );
return bias;
}
// Recompute gradients of hyperparameters and latent coordinates.
void GPCMMLPKernel::recompute(
const MatrixXd &gK, // Gradient of objective with respect to kernel.
const MatrixXd &gKd, // Gradient of objective with respect to diagonal kernel.
const MatrixXd* const *X, // Current latent positions.
MatrixXd **Xgrad // Latent position gradient.
)
{
assert(X[0] != NULL && X[1] == NULL); // Make sure we have only one element.
assert(Xgrad[1] == NULL); // Make sure we have only one element.
// Constants.
int N = gKd.rows();
// Compute gradient of variance.
vargrad(0,0) += kmat.cwiseProduct(gKd).sum();
// Compute gradient of weight and bias.
denominatorCubed.noalias() = denominator.array().pow(3).matrix();
baseCovGrad.noalias() = var(0,0)*gKd.cwiseQuotient((MatrixXd::Ones(arg.rows(),arg.cols()) - arg.cwiseProduct(arg)).cwiseSqrt());
vec.noalias() = innerProducts.diagonal();
wtgrad(0,0) += (innerProducts.cwiseQuotient(denominator) -
0.5*(numerator.cwiseQuotient(denominatorCubed)).cwiseProduct(
(wt(0,0)*vec + MatrixXd::Constant(vec.rows(),vec.cols(),bias(0,0)+1.0))*vec.transpose() +
vec*(wt(0,0)*vec + MatrixXd::Constant(vec.rows(),vec.cols(),bias(0,0)+1.0)).transpose())).cwiseProduct(baseCovGrad).sum();
biasgrad(0,0) += (denominator.array().inverse().matrix() -
0.5*(numerator.cwiseQuotient(denominatorCubed)).cwiseProduct(
(wt(0,0)*vec + MatrixXd::Constant(vec.rows(),vec.cols(),2.0*bias(0,0)+2.0)).replicate(1,vec.rows()) +
(wt(0,0)*vec.transpose()).replicate(vec.rows(),1))).cwiseProduct(baseCovGrad).sum();
// Compute X gradients.
if (Xgrad[0])
{
for (int d = 0; d < X[0]->cols(); d++)
{
MatrixXd b = (X[0]->rowwise().squaredNorm()*wt(0,0) + VectorXd::Constant(N,bias(0,0)+1.0)).replicate(1,N).cwiseProduct(
numerator).cwiseProduct((X[0]->col(d)*wt(0,0)*2.0).transpose().replicate(N,1)).cwiseQuotient(denominatorCubed);
Xgrad[0]->col(d) += ((X[0]->col(d)*(2.0*wt(0,0))).replicate(1,N).cwiseQuotient(denominator) -
((X[0]->rowwise().squaredNorm()*wt(0,0) + VectorXd::Constant(N,bias(0,0)+1.0)).replicate(1,N).cwiseProduct(
numerator).cwiseProduct((X[0]->col(d)*wt(0,0)*2.0).transpose().replicate(N,1)).cwiseQuotient(denominatorCubed))).cwiseProduct(
baseCovGrad).transpose().rowwise().sum();
}
}
}
PassRefPtr<FilterEffect> SVGFEConvolveMatrixElement::build(SVGFilterBuilder* filterBuilder)
{
FilterEffect* input1 = filterBuilder->getEffectById(in1());
if (!input1)
return 0;
Vector<float> kernelMatrixValues;
SVGNumberList* numbers = kernelMatrix();
ExceptionCode ec = 0;
int numberOfItems = numbers->numberOfItems();
for (int i = 0; i < numberOfItems; ++i)
kernelMatrixValues.append(numbers->getItem(i, ec));
int orderXValue = orderX();
int orderYValue = orderY();
if (!hasAttribute(SVGNames::orderAttr)) {
orderXValue = 3;
orderYValue = 3;
}
// The spec says this is a requirement, and should bail out if fails
if (orderXValue * orderYValue != numberOfItems)
return 0;
int targetXValue = targetX();
int targetYValue = targetY();
if (hasAttribute(SVGNames::targetXAttr) && (targetXValue < 0 || targetXValue >= orderXValue))
return 0;
// The spec says the default value is: targetX = floor ( orderX / 2 ))
if (!hasAttribute(SVGNames::targetXAttr))
targetXValue = static_cast<int>(floorf(orderXValue / 2));
if (hasAttribute(SVGNames::targetYAttr) && (targetYValue < 0 || targetYValue >= orderYValue))
return 0;
// The spec says the default value is: targetY = floor ( orderY / 2 ))
if (!hasAttribute(SVGNames::targetYAttr))
targetYValue = static_cast<int>(floorf(orderYValue / 2));
float divisorValue = divisor();
if (hasAttribute(SVGNames::divisorAttr) && !divisorValue)
return 0;
if (!hasAttribute(SVGNames::divisorAttr)) {
for (int i = 0; i < numberOfItems; ++i)
divisorValue += kernelMatrixValues[i];
if (!divisorValue)
divisorValue = 1;
}
RefPtr<FilterEffect> effect = FEConvolveMatrix::create(
IntSize(orderXValue, orderYValue), divisorValue,
bias(), IntPoint(targetXValue, targetYValue), static_cast<EdgeModeType>(edgeMode()),
FloatPoint(kernelUnitLengthX(), kernelUnitLengthX()), preserveAlpha(), kernelMatrixValues);
effect->inputEffects().append(input1);
return effect.release();
}
bvt float_utilst::build_constant(const ieee_floatt &src)
{
unbiased_floatt result;
result.sign=const_literal(src.get_sign());
result.NaN=const_literal(src.is_NaN());
result.infinity=const_literal(src.is_infinity());
result.exponent=bv_utils.build_constant(src.get_exponent(), spec.e);
result.fraction=bv_utils.build_constant(src.get_fraction(), spec.f+1);
return pack(bias(result));
}
inline float apply_float_controls(float value) const
{
value = (value - m_contrast_pivot) * m_contrast + m_contrast_pivot;
if (m_bias != 0.5f)
value = bias(value);
if (m_gain != 0.5f)
value = gain(value);
value = value * m_output_range + m_output_min;
if (m_clamp_min)
value = std::max(value, m_output_min);
if (m_clamp_max)
value = std::min(value, m_output_max);
return value;
}
//----------------------------------------------------------------------------//
void ScrollablePane::updateContainerPosition(void)
{
// basePos is the position represented by the scrollbars
// (these are negated so pane is scrolled in the correct directions)
UVector2 basePos(cegui_absdim(-getHorzScrollbar()->getScrollPosition()),
cegui_absdim(-getVertScrollbar()->getScrollPosition()));
// this bias is the absolute position that 0 on the scrollbars represent.
// Allows the pane to function correctly with negatively positioned content.
UVector2 bias(cegui_absdim(d_contentRect.d_min.d_x),
cegui_absdim(d_contentRect.d_min.d_y));
// set the new container pane position to be what the scrollbars request
// minus any bias generated by the location of the content.
getScrolledContainer()->setPosition(basePos - bias);
}
void testBivariateStats(){
float x[]={95.0,85.0,80.0,70.0,60.0};
float y[]={85.0,95.0,70.0,65.0,70.0};
regressionCoefficients coeffs;
coeffs=regression(x,y,5);
printf("\nBivariate Stats\n");
printf("correlation =%f\n",correlation(x,y,5));
printf("covariance =%f\n",covariance(x,y,5));
printf("rmse =%f\n",rmse(x,y,5));
printf("bias =%f\n",bias(x,y,5));
printf("m =%f\n",coeffs.m);
printf("c =%f\n",coeffs.c);
}
void ScrollablePane::updateContainerPosition(void)
{
assert(d_container != 0);
assert(d_horzScrollbar != 0);
assert(d_vertScrollbar != 0);
// basePos is the position represented by the scrollbars
// (these are negated so pane is scrolled in the correct directions)
Point basePos(-d_horzScrollbar->getScrollPosition(), -d_vertScrollbar->getScrollPosition());
// this bias is the absolute position that 0 on the scrollbars represent.
// effectively removes un-used empty space from the pane.
Point bias(d_contentRect.d_left, d_contentRect.d_top);
// set the new container pane position to be what the scrollbars request
// minus any bias generated by the location of the content.
d_container->setPosition(Absolute, basePos - bias);
}
// ---------------------------------------------------------------------------------------
//
void update()
{
float time = ofGetElapsedTimef();
ocl.begin();
cl::ImageGL cl_destImage;
ofxCL::convert( destImage.getTextureReference(), cl_destImage );
ofxCL::Kernel::Ref noiseKernel = noiseProgram.getKernel( methodNames.at(methodIndex) );
noiseKernel->global( destImage.getWidth(), destImage.getHeight() );
float t = time * 0.1f;
// ofParameters do not seem to get passed on correctly. Todo: avoid this workaround.
float tmpLacunarity = Lacunarity;
float tmpIncrement = Increment;
float tmpOctaves = Octaves;
float tmpAmplitude = Amplitude;
ofVec2f bias(0,0);
if( AutoMove )
{
bias.x = ofSignedNoise( t, t * -0.30f ) * AutoMoveMagnitude;
bias.y = ofSignedNoise( -t, t * 0.33f ) * AutoMoveMagnitude;
}
float tmpScale = Scale * ScaleMultiplier;
if( methodIndex == 0 )
{
noiseKernel->call( cl_destImage, bias, ofVec2f(tmpScale,tmpScale), tmpAmplitude );
}
else
{
noiseKernel->call( cl_destImage, bias, ofVec2f(tmpScale,tmpScale), tmpLacunarity, tmpIncrement, tmpOctaves, tmpAmplitude );
}
ocl.end();
}
unsigned int Forge::CalculateOddsRange(const SpecialInfo & specialInfo, Weapon weapon, const std::vector<int> & materials)
{
// Determine the bias
Bias bias(DetermineWeaponBias(specialInfo.special, weapon));
for (size_t i(0); i < materials.size(); ++i)
bias = ResolveBiases(bias, DetermineMaterialBias(specialInfo.special, materials[i]));
// Factor in the basic range for the special
int range(specialInfo.baseOdds);
switch (bias)
{
case Prohibited: return 0;
case Guaranteed: return GuaranteedRange;
case Likely: range = (range * 2) + 20; break;
case Unlikely: range = (range / 2) - 20; break;
default: break;
}
return static_cast<unsigned int>(UMAX(range, 0));
}
bvt float_utilst::rounder(const unbiased_floatt &src)
{
// incoming: some fraction (with explicit 1),
// some exponent without bias
// outgoing: rounded, with right size, with hidden bit, bias
bvt aligned_fraction=src.fraction,
aligned_exponent=src.exponent;
{
std::size_t exponent_bits=
std::max((std::size_t)integer2size_t(address_bits(spec.f)),
(std::size_t)spec.e)+1;
// before normalization, make sure exponent is large enough
if(aligned_exponent.size()<exponent_bits)
{
// sign extend
aligned_exponent=
bv_utils.sign_extension(aligned_exponent, exponent_bits);
}
}
// align it!
normalization_shift(aligned_fraction, aligned_exponent);
denormalization_shift(aligned_fraction, aligned_exponent);
unbiased_floatt result;
result.fraction=aligned_fraction;
result.exponent=aligned_exponent;
result.sign=src.sign;
result.NaN=src.NaN;
result.infinity=src.infinity;
round_fraction(result);
round_exponent(result);
return pack(bias(result));
}
// Write kernel data to file.
void GPCMMLPKernel::write(
GPCMMatWriter *writer // Writing interface.
)
{
GPCMKernel::write(writer); // Let superclass write first.
// Write parameters.
writer->writeDouble(wt(0,0),"weightVariance");
writer->writeDouble(bias(0,0),"biasVariance");
writer->writeDouble(var(0,0),"variance");
// Write priors.
GPCMMatWriter *cellWriter = writer->writeCell("priors",1,3);
GPCMMatWriter *priorStruct = cellWriter->writeStruct("",1,1);
priorStruct->writeDouble(1.0,"index");
wtprior->write(priorStruct);
priorStruct = cellWriter->closeStruct();
priorStruct = cellWriter->writeStruct("",1,1);
priorStruct->writeDouble(2.0,"index");
biasprior->write(priorStruct);
priorStruct = cellWriter->closeStruct();
priorStruct = cellWriter->writeStruct("",1,1);
priorStruct->writeDouble(3.0,"index");
varprior->write(priorStruct);
cellWriter->closeStruct();
writer->closeCell();
// Write transforms struct.
GPCMMatWriter *xformStruct = writer->writeStruct("transforms",1,3);
xformStruct->writeDouble(1.0,"index");
xformStruct->writeString("exp","type");
xformStruct = writer->closeStruct();
xformStruct->writeDouble(2.0,"index");
xformStruct->writeString("exp","type");
xformStruct = writer->closeStruct();
xformStruct->writeDouble(3.0,"index");
xformStruct->writeString("exp","type");
writer->closeStruct();
}
VectorXd LayeredFeedForwardNeuralNet::FireSingleLayer(const VectorXd& inputActivations, long layerIndex) const
{
// get layer input weights (also checks valid layerIndex)
const MatrixXd& layerInputWeights = GetLayerInputWeights(layerIndex);
if (layerInputWeights.cols() - 1 != inputActivations.size())
{
// input is invalid for this neural net topology
throw NeuralNetTopologyMismatch("activation input must match number of units in neural network layer");
}
// get the activation function
auto expressionParser = UnaryExpressionParserFactory::CreateParser();
UnaryFunction activationFunction = expressionParser->GetFunctionForExpression(m_activationFunction);
// bias activation
VectorXd bias(1);
bias << -1.0;
// calculate layer net inputs
VectorXd inputPlusBias(layerInputWeights.cols());
inputPlusBias << inputActivations, bias;
//std::cout << "layer " << layerIndex << " input activations +bias : " << std::endl << inputPlusBias << std::endl << std::endl;
//std::cout << "layer " << layerIndex << " input weights : " << std::endl << layerInputWeights << std::endl << std::endl;
VectorXd layerNetInputs = layerInputWeights * inputPlusBias;
//std::cout << "layer " << layerIndex << " net inputs : " << std::endl << layerNetInputs << std::endl << std::endl;
// calculate layer activations
VectorXd layerActivations = layerNetInputs.unaryExpr(activationFunction);
//std::cout << "layer " << layerIndex << " output activations : " << layerActivations << std::endl << std::endl;
return layerActivations;
}
int ColormapEdit::qt_metacall(QMetaObject::Call _c, int _id, void **_a)
{
_id = QDialog::qt_metacall(_c, _id, _a);
if (_id < 0)
return _id;
if (_c == QMetaObject::InvokeMetaMethod) {
switch (_id) {
case 0: newColorList((*reinterpret_cast< QList<QColor>(*)>(_a[1]))); break;
case 1: rotate((*reinterpret_cast< int(*)>(_a[1]))); break;
case 2: bias((*reinterpret_cast< int(*)>(_a[1]))); break;
case 3: contrast((*reinterpret_cast< int(*)>(_a[1]))); break;
case 4: setColor(); break;
case 5: okPressed(); break;
case 6: applyPressed(); break;
case 7: biasReset(); break;
case 8: rotateReset(); break;
case 9: contrastRest(); break;
case 10: updateColormap(); break;
case 11: resetMap((*reinterpret_cast< int(*)>(_a[1]))); break;
}
_id -= 12;
}
return _id;
}
// Return covariance of a single set of points.
MatrixXd GPCMMLPKernel::covariance(
const MatrixXd* const *X // Data matrix.
)
{
assert(X[0] != NULL && X[1] == NULL); // Make sure we have only one element.
innerProducts.noalias() = (*X[0])*X[0]->transpose();
numerator.noalias() = innerProducts*wt(0,0) + MatrixXd::Constant(X[0]->rows(),X[0]->rows(),bias(0,0));
vec.noalias() = numerator.diagonal() + VectorXd::Constant(X[0]->rows(),1.0);
denominator.noalias() = (vec*vec.transpose()).cwiseSqrt();
arg.noalias() = numerator.cwiseQuotient(denominator);
kmat.noalias() = arg.array().asin().matrix();
return var(0,0)*kmat;
}
void G1BiasedMappedArrayBase::verify_biased_index_inclusive_end(idx_t biased_index) const {
guarantee(_biased_base != NULL, "Array not initialized");
guarantee(biased_index >= bias() && biased_index <= (bias() + length()),
"Biased index out of inclusive bounds, index: " SIZE_FORMAT " bias: " SIZE_FORMAT " length: " SIZE_FORMAT,
biased_index, bias(), length());
}
bvt float_utilst::add_sub(
const bvt &src1,
const bvt &src2,
bool subtract)
{
unbiased_floatt unpacked1=unpack(src1);
unbiased_floatt unpacked2=unpack(src2);
// subtract?
if(subtract)
unpacked2.sign=!unpacked2.sign;
// figure out which operand has the bigger exponent
const bvt exponent_difference=subtract_exponents(unpacked1, unpacked2);
literalt src2_bigger=exponent_difference.back();
const bvt bigger_exponent=
bv_utils.select(src2_bigger, unpacked2.exponent, unpacked1.exponent);
// swap fractions as needed
const bvt new_fraction1=
bv_utils.select(src2_bigger, unpacked2.fraction, unpacked1.fraction);
const bvt new_fraction2=
bv_utils.select(src2_bigger, unpacked1.fraction, unpacked2.fraction);
// compute distance
const bvt distance=bv_utils.absolute_value(exponent_difference);
// limit the distance: shifting more than f+3 bits is unnecessary
const bvt limited_dist=limit_distance(distance, spec.f+3);
// pad fractions with 2 zeros from below
const bvt fraction1_padded=bv_utils.concatenate(bv_utils.zeros(3), new_fraction1);
const bvt fraction2_padded=bv_utils.concatenate(bv_utils.zeros(3), new_fraction2);
// shift new_fraction2
literalt sticky_bit;
const bvt fraction1_shifted=fraction1_padded;
const bvt fraction2_shifted=sticky_right_shift(
fraction2_padded, limited_dist, sticky_bit);
// sticky bit: or of the bits lost by the right-shift
bvt fraction2_stickied=fraction2_shifted;
fraction2_stickied[0]=prop.lor(fraction2_shifted[0], sticky_bit);
// need to have two extra fraction bits for addition and rounding
const bvt fraction1_ext=bv_utils.zero_extension(fraction1_shifted, fraction1_shifted.size()+2);
const bvt fraction2_ext=bv_utils.zero_extension(fraction2_stickied, fraction2_stickied.size()+2);
unbiased_floatt result;
// now add/sub them
literalt subtract_lit=prop.lxor(unpacked1.sign, unpacked2.sign);
result.fraction=
bv_utils.add_sub(fraction1_ext, fraction2_ext, subtract_lit);
// sign of result
literalt fraction_sign=result.fraction.back();
result.fraction=bv_utils.absolute_value(result.fraction);
result.exponent=bigger_exponent;
// adjust the exponent for the fact that we added two bits to the fraction
result.exponent=
bv_utils.add(bv_utils.sign_extension(result.exponent, result.exponent.size()+1),
bv_utils.build_constant(2, result.exponent.size()+1));
// NaN?
result.NaN=prop.lor(
prop.land(prop.land(unpacked1.infinity, unpacked2.infinity),
prop.lxor(unpacked1.sign, unpacked2.sign)),
prop.lor(unpacked1.NaN, unpacked2.NaN));
// infinity?
result.infinity=prop.land(
!result.NaN,
prop.lor(unpacked1.infinity, unpacked2.infinity));
// zero?
// Note that:
// 1. The zero flag isn't used apart from in divide and
// is only set on unpack
// 2. Subnormals mean that addition or subtraction can't round to 0,
// thus we can perform this test now
// 3. The rules for sign are different for zero
result.zero = prop.land(
!prop.lor(result.infinity, result.NaN),
!prop.lor(result.fraction));
// sign
literalt add_sub_sign=
prop.lxor(prop.lselect(src2_bigger, unpacked2.sign, unpacked1.sign),
fraction_sign);
literalt infinity_sign=
prop.lselect(unpacked1.infinity, unpacked1.sign, unpacked2.sign);
#if 1
literalt zero_sign=
prop.lselect(rounding_mode_bits.round_to_minus_inf,
prop.lor(unpacked1.sign, unpacked2.sign),
prop.land(unpacked1.sign, unpacked2.sign));
result.sign=prop.lselect(
result.infinity,
infinity_sign,
prop.lselect(result.zero,
zero_sign,
add_sub_sign));
#else
result.sign=prop.lselect(
result.infinity,
infinity_sign,
add_sub_sign);
#endif
#if 0
result.sign=const_literal(false);
result.fraction.resize(spec.f+1, const_literal(true));
result.exponent.resize(spec.e, const_literal(false));
result.NaN=const_literal(false);
result.infinity=const_literal(false);
//for(std::size_t i=0; i<result.fraction.size(); i++)
// result.fraction[i]=const_literal(true);
for(std::size_t i=0; i<result.fraction.size(); i++)
result.fraction[i]=new_fraction2[i];
return pack(bias(result));
#endif
return rounder(result);
}
/********************************************************
*
*
* Draw Robot
*
*
********************************************************/
void Virtual3dCharacterViewer::glDrawBody(Character3DBody& body)
{
glPushMatrix();
// getHipPos() ; // get Hip's Position from the IK calculator .
LocateHip (body);
body.HipPosInCamera = getCurTranslation () ;
body.HipPosInScreen = getScreenPos(body.HipPosInCamera) ;
RotateHip(body) ;
glPushMatrix(); // save the Hip Matrix
drawHip (body) ;
glTranslatef( 0.0, body.Model.LenHipToChest, 0.0 ) ;
RotateChest (body);
drawChest (body) ;
glTranslatef(0.0, body.Model.LenChestToNeck, 0.0);
body.NeckPosInCamera = getCurTranslation () ;
body.NeckPos = TransformVector( WorldMatInv, body.NeckPosInCamera ) ;
body.NeckPosInScreen = getScreenPos(body.NeckPosInCamera) ;
glPushMatrix(); // save the Neck Matrix
RotateLeftShoulder(body);
drawLeftShoulder(body);
glTranslatef(body.Model.LenNeckToShoulder, 0.0, 0.0);
RotateLeftUpperArm(body);
drawLeftUpperArm(body);
glTranslatef(body.Model.LenShoulderToElbow, 0.0, 0.0);
RotateLeftLowerArm(body);
drawLeftLowerArm(body);
glTranslatef(body.Model.LenElbowToWrist, 0.0, 0.0);
body.LeftWristPosInCamera = getCurTranslation () ;
Vector3D bias(-1,0,0) ;
body.LeftWristPos = TransformVector( WorldMatInv, body.LeftWristPosInCamera );
body.LeftWristPosInScreen = getScreenPos(body.LeftWristPosInCamera) ;
RotateLeftHand(body);
drawLeftHand(body);
glPopMatrix(); // reload the Neck Matrix
glPushMatrix(); // restore the Neck Matrix again .
RotateRightShoulder(body);
drawRightShoulder(body);
glTranslatef(-body.Model.LenNeckToShoulder, 0.0, 0.0);
RotateRightUpperArm(body);
drawRightUpperArm(body);
glTranslatef(-body.Model.LenShoulderToElbow, 0.0, 0.0);
RotateRightLowerArm(body);
drawRightLowerArm(body);
glTranslatef(-body.Model.LenElbowToWrist, 0.0, 0.0);
body.RightWristPosInCamera = getCurTranslation () ;
body.RightWristPos = TransformVector( WorldMatInv, body.RightWristPosInCamera ) ;
body.RightWristPosInScreen = getScreenPos(body.RightWristPosInCamera) ;
RotateRightHand(body);
drawRightHand(body);
glPopMatrix(); // reload the Neck Matrix again .
RotateHead(body);
drawHead (body);
glTranslatef(0.0, body.Model.LenNeckToHead, 0.0);
body.HeadPosInCamera = getCurTranslation () ;
body.HeadPos = TransformVector( WorldMatInv, body.HeadPosInCamera ) ;
body.HeadPosInScreen = getScreenPos(body.HeadPosInCamera) ;
glPopMatrix(); // reload the Hip Matrix .
glPushMatrix(); // restore the Hip Matrix again.
RotateLeftThighRoot (body);
drawLeftThighRoot(body);
glTranslatef(body.Model.LenHipToThigh, 0.0, 0.0);
RotateLeftThigh (body) ;
drawLeftThigh(body);
glTranslatef(0.0, -body.Model.LenThighToKnee, 0.0);
RotateLeftShank(body);
drawLeftShank(body);
glTranslatef(0.0, -body.Model.LenKneeToAnkle, 0.0);
body.LeftAnklePosInCamera = getCurTranslation();
body.LeftAnklePos = TransformVector( WorldMatInv, body.LeftAnklePosInCamera ) ;
body.LeftAnklePosInScreen = getScreenPos(body.LeftAnklePosInCamera) ;
RotateLeftFoot (body);
drawLeftFoot(body);
glPopMatrix(); // reload the Hip Matrix again.
RotateRightThighRoot (body);
drawRightThighRoot(body);
glTranslatef(-body.Model.LenHipToThigh, 0.0, 0.0);
RotateRightThigh (body) ;
drawRightThigh(body);
glTranslatef(0.0, -body.Model.LenThighToKnee, 0.0);
RotateRightShank(body);
drawRightShank(body);
glTranslatef(0.0, -body.Model.LenKneeToAnkle, 0.0);
body.RightAnklePosInCamera = getCurTranslation();
body.RightAnklePos = TransformVector( WorldMatInv, body.RightAnklePosInCamera ) ;
body.RightAnklePosInScreen = getScreenPos(body.RightAnklePosInCamera) ;
RotateRightFoot (body);
drawRightFoot(body);
glPopMatrix();
}
void MD5::prepModel(void)
{
// Prepare Buffer Objects
int offset = 0;
size_t vboSize = sizeof(glm::vec2) + sizeof(glm::vec4) + sizeof(glm::mat4);
glGenBuffers(1, &vbo);
glBindBuffer(GL_ARRAY_BUFFER, vbo);
glBufferData(GL_ARRAY_BUFFER, vboSize*numVerts, NULL, GL_STATIC_DRAW);
// Generate Vertex Buffer Object
for(int i = 0; i < numMeshes; ++i)
{
for(int j = 0; j < meshList[i].getNumVert(); ++j)
{
Mesh::Vertex v = meshList[i].getVerts(j);
// Add UV coords
glBufferSubData(GL_ARRAY_BUFFER, offset, sizeof(glm::vec2), glm::value_ptr(v.uv));
offset += sizeof(glm::vec2);
// Add vertex weights
glm::vec4 bias(0.0f);
glm::mat4 pos(0.0f);
int index = v.weightIndex;
for(int k = 0; k < v.weightElem; ++k)
{
Mesh::Weight w = meshList[i].getWeight(index+k);
bias[k] = w.value;
glm::vec4 posi(0, 0, 0, -1);
posi.x = w.position.x;
posi.y = w.position.y;
posi.z = w.position.z;
posi.w = w.jointIndex;
pos[k] = posi;
}
glBufferSubData(GL_ARRAY_BUFFER, offset, sizeof(glm::vec4), glm::value_ptr(bias));
offset += sizeof(glm::vec4);
glBufferSubData(GL_ARRAY_BUFFER, offset, sizeof(glm::mat4), glm::value_ptr(pos));
offset += sizeof(glm::mat4) + (sizeof(float)*2);
}
}
offset = 0;
// Generate Index Buffer Object
glGenBuffers(1, &ibo);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, ibo);
glBufferData(GL_ELEMENT_ARRAY_BUFFER, sizeof(GLushort)*3*numTris, NULL, GL_STATIC_DRAW);
for(int i = 0; i < numMeshes; ++i)
{
for(int j = 0; j < meshList[i].getNumTri(); ++j)
{
Mesh::Triangle t = meshList[i].getTris(j);
t.v1 += offset;
t.v2 += offset;
t.v3 += offset;
glBufferSubData(GL_ELEMENT_ARRAY_BUFFER, offset, sizeof(GLushort)*3, (GLvoid*)&t);
}
offset += meshList[i].getNumTri();
}
GLuint uvLoc = 0;
GLuint wBias = 1;
GLuint wPos = 2;
if(GLEW_ARB_uniform_buffer_object)
{
glBindAttribLocation(shaderProgram, 0, "uv");
glBindAttribLocation(shaderProgram, 1, "wBias");
glBindAttribLocation(shaderProgram, 2, "wPos");
}
else
{
uvLoc = glGetAttribLocation(shaderProgram, "uv");
wBias = glGetAttribLocation(shaderProgram, "wBias");
wPos = glGetAttribLocation(shaderProgram, "wPos");
}
// Generate and bind Vertex Array Object
glGenVertexArrays(1, &vao);
glBindVertexArray(vao);
//Enable and link attributes
glEnableVertexAttribArray(uvLoc);
glEnableVertexAttribArray(wBias);
glEnableVertexAttribArray(wPos);
glEnableVertexAttribArray(wPos+1);
glEnableVertexAttribArray(wPos+2);
glEnableVertexAttribArray(wPos+3);
glBindBuffer(GL_ARRAY_BUFFER, vbo);
glVertexAttribPointer(uvLoc, 2, GL_FLOAT, GL_FALSE, vboSize, 0);
glVertexAttribPointer(wBias, 4, GL_FLOAT, GL_FALSE, vboSize, (GLvoid*)(sizeof(glm::vec2)));
for(int i = 0; i < 4; ++i)
{
glVertexAttribPointer(wPos+i, 4, GL_FLOAT, GL_FALSE, vboSize, (GLvoid*)(sizeof(glm::vec2)+sizeof(glm::vec4)+(sizeof(glm::vec4)*i)));
}
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, ibo);
glBindVertexArray(0);
}
// Return coviarance of two sets of points.
MatrixXd GPCMMLPKernel::covariance(
const MatrixXd* const *X1, // First data matrix.
const MatrixXd* const *X2 // Second data matrix.
)
{
assert(X1[0] != NULL && X1[1] == NULL); // Make sure we have only one element.
assert(X2[0] != NULL && X2[1] == NULL); // Make sure we have only one element.
return var(0,0)*(((((*X1[0])*X2[0]->transpose())*wt(0,0) + MatrixXd::Constant(X1[0]->rows(),X1[0]->rows(),bias(0,0))).cwiseQuotient(
((X1[0]->rowwise().squaredNorm()*wt(0,0) + MatrixXd::Constant(X1[0]->rows(),1,bias(0,0)+1.0))*
(X2[0]->rowwise().squaredNorm()*wt(0,0) + MatrixXd::Constant(X1[0]->rows(),1,bias(0,0)+1.0)).transpose()).cwiseSqrt())).array().asin().matrix());
}
void _jit_avx512_core_fp32_wino_conv_4x3_t<is_fwd>::_execute_data_W_SGD(
float *inp_ptr, float *out_ptr, float *wei_ptr, float *bias_ptr,
const memory_tracking::grantor_t &scratchpad) const {
const auto &jcp = kernel_->jcp;
const auto &p_ops = attr_->post_ops_;
const int inph = is_fwd ? jcp.ih : jcp.oh;
const int inpw = is_fwd ? jcp.iw : jcp.ow;
const int outh = is_fwd ? jcp.oh : jcp.ih;
const int outw = is_fwd ? jcp.ow : jcp.iw;
array_offset_calculator<float, 5> input(inp_ptr,
jcp.mb, jcp.dimK/jcp.dimK_reg_block, inph, inpw, jcp.dimK_reg_block);
array_offset_calculator<float, 5> output(out_ptr,
jcp.mb, jcp.dimM/jcp.dimM_simd_block, outh, outw, jcp.dimM_simd_block);
array_offset_calculator<float, 6> weights(wei_ptr,
jcp.oc/jcp.oc_simd_block, jcp.ic/jcp.ic_simd_block, jcp.kh, jcp.kw,
jcp.ic_simd_block, jcp.oc_simd_block);
array_offset_calculator<float, 2> bias(bias_ptr,
jcp.oc/jcp.oc_simd_block, jcp.oc_simd_block);
auto wino_wei = (jcp.prop_kind == prop_kind::forward_inference)
? wei_ptr
: scratchpad.template get<float>(key_wino_U);
array_offset_calculator<float, 8> U(wino_wei,
jcp.dimM_nb_block,
alpha, alpha,
jcp.dimK_nb_block,
jcp.dimM_block * jcp.dimM_reg_block, jcp.dimK_block,
jcp.dimK_reg_block, jcp.dimM_simd_block);
array_offset_calculator<float, 8> M(is_fwd
? scratchpad.template get<float>(key_wino_M)
: scratchpad.template get<float>(key_wino_V),
0, jcp.dimM_nb_block, alpha, alpha,
jcp.dimN_block, jcp.dimM_block * jcp.dimM_reg_block,
jcp.dimN_reg_block, jcp.dimM_simd_block);
array_offset_calculator<float, 8> V(is_fwd
? scratchpad.template get<float>(key_wino_V)
: scratchpad.template get<float>(key_wino_M),
0, alpha, alpha, jcp.dimN_block,
jcp.dimK_nb_block, jcp.dimK_block,
jcp.dimN_reg_block, jcp.dimK_reg_block);
const bool wants_padded_bias = jcp.with_bias
&& jcp.oc_without_padding != jcp.oc;
float last_slice_bias[simd_w] = {0};
if (wants_padded_bias) {
for (int oc = 0; oc < jcp.oc_without_padding % jcp.oc_simd_block; ++oc)
last_slice_bias[oc] = bias(jcp.dimM / jcp.dimM_simd_block - 1, oc);
}
if (jcp.prop_kind != prop_kind::forward_inference) {
parallel_nd(jcp.nb_oc, jcp.nb_ic, (jcp.oc_block * jcp.oc_reg_block), (jcp.ic_block * jcp.ic_reg_block),
[&](int ofm1, int ifm1, int ofm2, int ifm2) {
float *U_base_ptr = is_fwd
? &(U(ofm1, 0, 0, ifm1, ofm2, ifm2, 0, 0))
: &(U(ifm1, 0, 0, ofm1, ifm2, ofm2, 0, 0));
weight_transform_data(jcp,
&(weights(
ofm1 * jcp.oc_block * jcp.oc_reg_block + ofm2,
ifm1 * jcp.ic_block * jcp.ic_reg_block + ifm2,
0, 0, 0, 0)),
U_base_ptr);
});
}
PRAGMA_OMP(parallel)
{
int ithr = mkldnn_get_thread_num();
PRAGMA_OMP(for schedule(static))
for (int tile_block = 0; tile_block < jcp.tile_block; tile_block++) {
for (int K_blk1 = 0; K_blk1 < jcp.dimK_nb_block; K_blk1++) {
for (int K_blk2 = 0; K_blk2 < jcp.dimK_block; K_blk2++) {
input_transform_tileblock_data(
tile_block, jcp,
&(input(0, K_blk1 * jcp.dimK_block + K_blk2, 0, 0, 0)),
&(V(ithr, 0, 0, 0, K_blk1, K_blk2, 0, 0)));
}
}
for (int oj = 0; oj < alpha; oj++) {
for (int oi = 0; oi < alpha; oi++) {
for (int M_blk1 = 0; M_blk1 < jcp.dimM_nb_block; M_blk1++)
for (int K_blk1 = 0; K_blk1 < jcp.dimK_nb_block; K_blk1++)
for (int N_blk = 0; N_blk < jcp.dimN_block; N_blk++)
kernel_->gemm_loop_ker(
(float *)&(M(ithr, M_blk1, oj, oi,
N_blk, 0, 0, 0)),
(const float *)&(U(M_blk1, oj, oi, K_blk1,
0, 0, 0, 0)),
(const float *)&(V(ithr, oj, oi,
N_blk, K_blk1, 0, 0, 0)), K_blk1);
}
}
for (int M_blk1 = 0; M_blk1 < jcp.dimM_nb_block; M_blk1++) {
for (int M_blk2 = 0; M_blk2 < jcp.dimM_block * jcp.dimM_reg_block;
M_blk2++) {
const int M_blk =
M_blk1 * jcp.dimM_block * jcp.dimM_reg_block + M_blk2;
float *bias_ptr = wants_padded_bias
&& M_blk == jcp.dimM / jcp.dimM_simd_block - 1
? last_slice_bias : &bias(M_blk, 0);
output_transform_tileblock_data(tile_block, jcp, p_ops,
&(M(ithr, M_blk1, 0, 0, 0, M_blk2, 0, 0)),
&(output(0, M_blk, 0, 0, 0)), bias_ptr);
}
}
}
}
}
void _jit_avx512_core_fp32_wino_conv_4x3_t<is_fwd>::_execute_data_W_S_G_D(
float *inp_ptr, float *out_ptr, float *wei_ptr, float *bias_ptr,
const memory_tracking::grantor_t &scratchpad) const {
const auto &jcp = kernel_->jcp;
const auto &p_ops = attr_->post_ops_;
const int inph = is_fwd ? jcp.ih : jcp.oh;
const int inpw = is_fwd ? jcp.iw : jcp.ow;
const int outh = is_fwd ? jcp.oh : jcp.ih;
const int outw = is_fwd ? jcp.ow : jcp.iw;
/* Notation:
FWD: dimM:oc, dimN:ntiles, dimK:ic,
BWD: dimM:ic, dimN:ntiles, dimK:oc,
FWD/BWD: V: src/diff_dst transform, U:weight transform,
M:dst/diff_src transform */
array_offset_calculator<float, 5> input(inp_ptr,
jcp.mb, jcp.dimK/jcp.dimK_reg_block, inph, inpw,
jcp.dimK_reg_block);
array_offset_calculator<float, 5> output(out_ptr,
jcp.mb, jcp.dimM/jcp.dimM_simd_block, outh, outw,
jcp.dimM_simd_block);
array_offset_calculator<float, 6> weights(wei_ptr,
jcp.oc/jcp.oc_simd_block, jcp.ic/jcp.ic_simd_block, jcp.kh, jcp.kw,
jcp.ic_simd_block, jcp.oc_simd_block);
array_offset_calculator<float, 2> bias(bias_ptr,
jcp.dimM/jcp.dimM_simd_block, jcp.dimM_simd_block);
array_offset_calculator<float, 8> M(is_fwd
? scratchpad.template get<float>(key_wino_M)
: scratchpad.template get<float>(key_wino_V),
jcp.dimN_nb_block, jcp.dimM_nb_block,
alpha, alpha,
jcp.dimN_block, jcp.dimM_block * jcp.dimM_reg_block,
jcp.dimN_reg_block, jcp.dimM_simd_block);
auto wino_wei = (jcp.prop_kind == prop_kind::forward_inference)
? wei_ptr
: scratchpad.template get<float>(key_wino_U);
array_offset_calculator<float, 8> U(wino_wei,
jcp.dimM_nb_block,
alpha, alpha,
jcp.dimK_nb_block,
jcp.dimM_block * jcp.dimM_reg_block, jcp.dimK_block,
jcp.dimK_reg_block, jcp.dimM_simd_block);
array_offset_calculator<float, 8> V(is_fwd
? scratchpad.template get<float>(key_wino_V)
: scratchpad.template get<float>(key_wino_M),
jcp.dimN_nb_block, alpha, alpha,
jcp.dimN_block, jcp.dimK_nb_block,
jcp.dimK_block, jcp.dimN_reg_block, jcp.dimK_reg_block);
const bool wants_padded_bias = jcp.with_bias
&& jcp.oc_without_padding != jcp.oc;
float last_slice_bias[simd_w] = {0};
if (wants_padded_bias) {
for (int oc = 0; oc < jcp.oc_without_padding % jcp.oc_simd_block; ++oc)
last_slice_bias[oc] = bias(jcp.dimM / jcp.dimM_simd_block - 1, oc);
}
PRAGMA_OMP(parallel)
{
parallel_nd_in_omp(jcp.mb, jcp.dimK_nb_block, jcp.dimK_block,
[&](int img, int K_blk1, int K_blk2) {
input_transform_data(img, jcp,
&(input(img, K_blk1 * jcp.dimK_block + K_blk2,
0, 0, 0)),
&(V(0, 0, 0, 0, K_blk1, K_blk2, 0, 0)));
});
if (jcp.prop_kind != prop_kind::forward_inference) {
parallel_nd_in_omp(jcp.nb_oc, jcp.nb_ic, (jcp.oc_block * jcp.oc_reg_block),
(jcp.ic_block * jcp.ic_reg_block),
[&](int ofm1, int ifm1, int ofm2, int ifm2) {
float *U_base_ptr = is_fwd
? &(U(ofm1, 0, 0, ifm1, ofm2, ifm2, 0, 0))
: &(U(ifm1, 0, 0, ofm1, ifm2, ofm2, 0, 0));
weight_transform_data(jcp,
&(weights(
ofm1 * jcp.oc_block * jcp.oc_reg_block + ofm2,
ifm1 * jcp.ic_block * jcp.ic_reg_block + ifm2,
0, 0, 0, 0)),
U_base_ptr);
});
}
PRAGMA_OMP(barrier)
parallel_nd_in_omp(jcp.dimN_nb_block, alpha, alpha, jcp.dimM_nb_block,
[&](int N_blk1, int oj, int oi, int M_blk1) {
for (int K_blk1 = 0; K_blk1 < jcp.dimK_nb_block;
K_blk1++)
for (int N_blk2 = 0; N_blk2 < jcp.dimN_block; N_blk2++)
kernel_->gemm_loop_ker(
(float *)&(M(N_blk1, M_blk1, oj, oi,
N_blk2, 0, 0, 0)),
(const float *)&(U(M_blk1, oj, oi,
K_blk1, 0, 0, 0, 0)),
(const float *)&(V(N_blk1, oj, oi,
N_blk2, K_blk1, 0, 0, 0)), K_blk1);
});
PRAGMA_OMP(barrier)
parallel_nd_in_omp(jcp.mb, jcp.dimM_nb_block, (jcp.dimM_block * jcp.dimM_reg_block),
[&](int img, int M_blk1, int M_blk2) {
const int M_blk =
M_blk1 * jcp.dimM_block * jcp.dimM_reg_block + M_blk2;
float *bias_ptr = wants_padded_bias
&& M_blk == jcp.dimM / jcp.dimM_simd_block - 1
? last_slice_bias : &bias(M_blk, 0);
output_transform_data(img, jcp, p_ops,
&(M(0, M_blk1, 0, 0, 0, M_blk2, 0, 0)),
&(output(img, M_blk, 0, 0, 0)), bias_ptr);
});
}
}
void ShadowMap::setShaderArgsRead(UniformTable& args, const String& prefix) const {
// The notNull macro is set by the depthTexture()
args.setUniform(prefix + "MVP", unitLightMVP());
args.setUniform(prefix + "bias", bias());
depthTexture()->setShaderArgs(args, prefix, Sampler::shadow());
}
void ofxVectorField::bias(float amt) {
bias(amt, amt);
}
/* Just like the halo mass function, we'll set this up such that
* you just need to interpolate.
*
* Now this has been changed to calculate the spatial-scale dependence
* of the bias factor. If you don't want the scale dep. b, then just
* input r<0.
*
* If the global flag LINEAR_PSP==0, uses the scale dependence calculated for
* for halo bias relative to the non-linear matter \xi_m(r):
*
* f^2(r) = (1.0+xi*1.17)^1.49/(1.0+xi*0.69)^2.09 --> b(r) = b0*f(r)
*
* For LINEAR_PSP==1, use scale dependence determined for the linear P(k):
*
* f(r) = 1 + exp[-(r/A)^0.7] --> where A is a parameter that we're gonna have to
* determine in more detail, but now A \approx 1
*/
double bias_interp(double m, double r)
{
static int flag=0,prev_cosmo=0, n;
static double *x,*y,*y2, pnorm;
int i;
double dm,max=16.3,min=9,a,b,m1,m2,dm1,xi,power,rm,sig,b1,b2,mass,rvir,a1;
double c1,c2,d1,d2;
if(!flag || RESET_COSMOLOGY!=prev_cosmo)
{
n = 100;
if(!ThisTask && OUTPUT)
fprintf(stdout,"RESET: resetting bias for %f %f\n",OMEGA_M,SIGMA_8);
if(!flag)
{
x=dvector(1,n);
y=dvector(1,n);
y2=dvector(1,n);
}
flag=1;
dm=(double)(max-min)/n;
for(i=1;i<=n;++i)
{
x[i]=pow(10.0,min+i*dm);
y[i]=log(bias(x[i]));
//printf("BIAS %e %e\n",x[i], exp(y[i]));
if(isinf(y[i])){ n = i-1; break; }
if(isnan(y[i])){ n = 1-1; break; }
x[i] = log(x[i]);
continue;
// no longer need to do this part, since we're taking into account
// halo overdensity in the bias formula.
/*
if(DELTA_HALO!=200)
{
x[i]=log(halo_mass_conversion2(x[i],halo_c200(x[i]),200.0,DELTA_HALO));
}
else
{
x[i]=log(x[i]);
}
*/
}
spline(x,y,n,2.0E+30,2.0E+30,y2);
prev_cosmo=RESET_COSMOLOGY;
pnorm=SIGMA_8/sigmac(8.0);
}
m=log(m);
splint(x,y,y2,n,m,&a);
a = exp(a);
// if we're using systematic errors in an MCMC, adjust parameter a1 (amplitude)
if(USE_ERRORS)
a *= M2N.bias_amp;
// if no scale-dependence required, return the large-scale value
if(r<0) return a;
/*
* SCALE DEPENDENT BIAS!!!
*----------------------------------------------------------------------
*/
/* FOR M/N analysis, use the Tinker etal 2005 result:
*/
xi = xi_interp(r);
if(M2N.scalebias_selfcal)
{
b = pow(1.0+xi*M2N.bias_amp1,M2N.bias_pow1)*pow(1.0+xi*0.69,-1.045);
}
else
{
//Attempting to correct scale-dep bias issue
//SCALE-DEP BIAS CORRECTION
rvir = pow(3*exp(m)/(4*PI*200*RHO_CRIT*OMEGA_M),THIRD);
// if(r<2.4*rvir)r=2.4*rvir;
if(r<2.8*rvir)r=2.8*rvir;
//rvir = pow(3*exp(m)/(4*PI*DELTA_HALO*RHO_CRIT*OMEGA_M),THIRD);
//if(r<2*rvir)r=2*rvir;
xi = xi_interp(r);
//parameters for scale-dep bias
//original values
//c1 = 1.17;
//c2 = 0.69;
//d1 = 1.49;
//d2 = -2.09;
c1 = HBIAS_C1;
c2 = HBIAS_C2;
d1 = HBIAS_D1;
d2 = (-1.)*HBIAS_D2;
/**
c1 = 1.52;
c2 = 0.84;
d1 = 1.49;
d2 = -2.09;
**/
b = pow(1.0+xi*c1,d1/2.)*pow(1.0+xi*c2,d2/2.0);
// if we're using systematic errors in an MCMC, adjust parameter a1 (amplitude)
if(USE_ERRORS) {
b = (b-1)*M2N.scalebias_amp + 1;
if(b<0)b=0;
}
}
return b*a;
/* first, calculate the standard Zheng-like, mass-independent scale bias
* based on the non-linear correlation function
* (re-calibrated using the SO catalogs)
*/
xi = xi_interp(r);
b = pow(1.0+xi*0.92,2.08)*pow(1.0+xi*0.74,-2.37);
if(b<0.6)b=0.6;
/* Now the mass-dependent term.
* Where the mass-dependence comes in the form of the large-scale bias itself
*/
sig = sigmac_radius_interp(r);
//if(a<0)
//b *= pow(1 + 0.028*pow(a,1.53)*pow(sig,1.57),2.59)/
//pow(1 + 0.253*pow(a,1.24)*pow(sig,1.71),0.397);
// if we're using systematic errors in an MCMC, adjust parameter a1 (amplitude)
if(USE_ERRORS) {
b = (b-1)*M2N.scalebias_amp + 1;
if(b<0)b=0;
}
return b*a;
}
void ONNXImporter::populateNet(Net dstNet)
{
CV_Assert(model_proto.has_graph());
opencv_onnx::GraphProto graph_proto = model_proto.graph();
std::map<std::string, Mat> constBlobs = getGraphTensors(graph_proto);
// List of internal blobs shapes.
std::map<std::string, MatShape> outShapes;
// Add all the inputs shapes. It includes as constant blobs as network's inputs shapes.
for (int i = 0; i < graph_proto.input_size(); ++i)
{
opencv_onnx::ValueInfoProto valueInfoProto = graph_proto.input(i);
CV_Assert(valueInfoProto.has_type());
opencv_onnx::TypeProto typeProto = valueInfoProto.type();
CV_Assert(typeProto.has_tensor_type());
opencv_onnx::TypeProto::Tensor tensor = typeProto.tensor_type();
CV_Assert(tensor.has_shape());
opencv_onnx::TensorShapeProto tensorShape = tensor.shape();
MatShape inpShape(tensorShape.dim_size());
for (int j = 0; j < inpShape.size(); ++j)
{
inpShape[j] = tensorShape.dim(j).dim_value();
}
outShapes[valueInfoProto.name()] = inpShape;
}
std::string framework_name;
if (model_proto.has_producer_name()) {
framework_name = model_proto.producer_name();
}
// create map with network inputs (without const blobs)
std::map<std::string, LayerInfo> layer_id;
std::map<std::string, LayerInfo>::iterator layerId;
std::map<std::string, MatShape>::iterator shapeIt;
// fill map: push layer name, layer id and output id
std::vector<String> netInputs;
for (int j = 0; j < graph_proto.input_size(); j++)
{
const std::string& name = graph_proto.input(j).name();
if (constBlobs.find(name) == constBlobs.end()) {
netInputs.push_back(name);
layer_id.insert(std::make_pair(name, LayerInfo(0, netInputs.size() - 1)));
}
}
dstNet.setInputsNames(netInputs);
int layersSize = graph_proto.node_size();
LayerParams layerParams;
opencv_onnx::NodeProto node_proto;
for(int li = 0; li < layersSize; li++)
{
node_proto = graph_proto.node(li);
layerParams = getLayerParams(node_proto);
CV_Assert(node_proto.output_size() >= 1);
layerParams.name = node_proto.output(0);
std::string layer_type = node_proto.op_type();
layerParams.type = layer_type;
if (layer_type == "MaxPool")
{
layerParams.type = "Pooling";
layerParams.set("pool", "MAX");
layerParams.set("ceil_mode", isCeilMode(layerParams));
}
else if (layer_type == "AveragePool")
{
layerParams.type = "Pooling";
layerParams.set("pool", "AVE");
layerParams.set("ceil_mode", isCeilMode(layerParams));
layerParams.set("ave_pool_padded_area", framework_name == "pytorch");
}
else if (layer_type == "GlobalAveragePool")
{
layerParams.type = "Pooling";
layerParams.set("pool", "AVE");
layerParams.set("global_pooling", true);
}
else if (layer_type == "Add" || layer_type == "Sum")
{
if (layer_id.find(node_proto.input(1)) == layer_id.end())
{
Mat blob = getBlob(node_proto, constBlobs, 1);
blob = blob.reshape(1, 1);
if (blob.total() == 1) {
layerParams.type = "Power";
layerParams.set("shift", blob.at<float>(0));
}
else {
layerParams.type = "Scale";
layerParams.set("bias_term", true);
layerParams.blobs.push_back(blob);
}
}
else {
layerParams.type = "Eltwise";
}
}
else if (layer_type == "Sub")
{
Mat blob = getBlob(node_proto, constBlobs, 1);
if (blob.total() == 1) {
layerParams.type = "Power";
layerParams.set("shift", -blob.at<float>(0));
}
else {
layerParams.type = "Scale";
layerParams.set("has_bias", true);
layerParams.blobs.push_back(-1.0f * blob.reshape(1, 1));
}
}
else if (layer_type == "Div")
{
Mat blob = getBlob(node_proto, constBlobs, 1);
CV_Assert_N(blob.type() == CV_32F, blob.total());
if (blob.total() == 1)
{
layerParams.set("scale", 1.0f / blob.at<float>(0));
layerParams.type = "Power";
}
else
{
layerParams.type = "Scale";
divide(1.0, blob, blob);
layerParams.blobs.push_back(blob);
layerParams.set("bias_term", false);
}
}
else if (layer_type == "Constant")
{
CV_Assert(node_proto.input_size() == 0);
CV_Assert(layerParams.blobs.size() == 1);
constBlobs.insert(std::make_pair(layerParams.name, layerParams.blobs[0]));
continue;
}
else if (layer_type == "ImageScaler")
{
const float scale = layerParams.has("scale") ? layerParams.get<float>("scale") : 1.0f;
layerParams.erase("scale");
if (layerParams.has("bias"))
{
layerParams.type = "Scale";
layerParams.blobs.push_back(
Mat(Size(1, layerParams.get("bias").size()), CV_32FC1, scale));
layerParams.set("bias_term", true);
Mat bias(1, layerParams.get("bias").size(), CV_32FC1);
for (int j = 0; j < bias.total(); j++) {
bias.at<float>(0, j) = layerParams.get("bias").getRealValue(j);
}
layerParams.blobs.push_back(bias);
layerParams.erase("bias");
}
else {
layerParams.set("scale", scale);
layerParams.type = "Power";
}
}
else if (layer_type == "LeakyRelu")
{
layerParams.type = "ReLU";
replaceLayerParam(layerParams, "alpha", "negative_slope");
}
else if (layer_type == "LRN")
{
replaceLayerParam(layerParams, "size", "local_size");
}
else if (layer_type == "BatchNormalization")
{
if (node_proto.input_size() != 5)
CV_Error(Error::StsNotImplemented,
"Expected input, scale, bias, mean and var");
layerParams.type = "BatchNorm";
replaceLayerParam(layerParams, "epsilon", "eps");
replaceLayerParam(layerParams, "spatial", "use_global_stats");
Mat meanData = getBlob(node_proto, constBlobs, 3);
Mat stdData = getBlob(node_proto, constBlobs, 4);
layerParams.blobs.push_back(meanData);
layerParams.blobs.push_back(stdData);
if (!node_proto.input(1).empty()) {
layerParams.set("has_weight", true);
layerParams.blobs.push_back(getBlob(node_proto, constBlobs, 1)); // weightData
} else {
layerParams.set("has_weight", false);
}
if (!node_proto.input(2).empty()) {
layerParams.set("has_bias", true);
layerParams.blobs.push_back(getBlob(node_proto, constBlobs, 2)); // biasData
} else {
layerParams.set("has_bias", false);
}
}
else if (layer_type == "Gemm")
{
CV_Assert(node_proto.input_size() >= 2);
layerParams.type = "InnerProduct";
Mat weights = getBlob(node_proto, constBlobs, 1);
int ind_num_out = 0;
if (layerParams.has("transB") && !layerParams.get<int>("transB")) {
transpose(weights, weights);
ind_num_out = 1;
}
layerParams.blobs.push_back(weights);
if (node_proto.input_size() == 3) {
Mat bias = getBlob(node_proto, constBlobs, 2);
layerParams.blobs.push_back(bias);
}
layerParams.set("num_output", layerParams.blobs[0].size[ind_num_out]);
layerParams.set("bias_term", node_proto.input_size() == 3);
}
else if (layer_type == "MatMul")
{
CV_Assert(node_proto.input_size() == 2);
layerParams.type = "InnerProduct";
Mat blob = getBlob(node_proto, constBlobs, 1);
layerParams.blobs.push_back(blob.t());
layerParams.set("bias_term", false);
layerParams.set("num_output", layerParams.blobs[0].size[0]);
}
else if (layer_type == "Mul")
{
CV_Assert(node_proto.input_size() == 2);
if (layer_id.find(node_proto.input(1)) == layer_id.end()) {
Mat blob = getBlob(node_proto, constBlobs, 1);
blob = blob.reshape(1, 1);
if (blob.total() == 1) {
layerParams.set("scale", blob.at<float>(0));
layerParams.type = "Power";
}
else {
layerParams.blobs.push_back(blob);
layerParams.type = "Scale";
}
}
else {
layerParams.type = "Eltwise";
layerParams.set("operation", "prod");
}
}
else if (layer_type == "Conv")
{
CV_Assert(node_proto.input_size() >= 2);
layerParams.type = "Convolution";
for (int j = 1; j < node_proto.input_size(); j++) {
layerParams.blobs.push_back(getBlob(node_proto, constBlobs, j));
}
layerParams.set("num_output", layerParams.blobs[0].size[0]);
layerParams.set("bias_term", node_proto.input_size() == 3);
}
else if (layer_type == "ConvTranspose")
{
CV_Assert(node_proto.input_size() >= 2);
layerParams.type = "Deconvolution";
for (int j = 1; j < node_proto.input_size(); j++) {
layerParams.blobs.push_back(getBlob(node_proto, constBlobs, j));
}
layerParams.set("num_output", layerParams.blobs[0].size[1]);
layerParams.set("bias_term", node_proto.input_size() == 3);
}
else if (layer_type == "Transpose")
{
layerParams.type = "Permute";
replaceLayerParam(layerParams, "perm", "order");
}
else if (layer_type == "Unsqueeze")
{
CV_Assert(node_proto.input_size() == 1);
Mat input = getBlob(node_proto, constBlobs, 0);
DictValue axes = layerParams.get("axes");
std::vector<int> dims;
for (int j = 0; j < input.dims; j++) {
dims.push_back(input.size[j]);
}
CV_Assert(axes.getIntValue(axes.size()-1) <= dims.size());
for (int j = 0; j < axes.size(); j++) {
dims.insert(dims.begin() + axes.getIntValue(j), 1);
}
Mat out = input.reshape(0, dims);
constBlobs.insert(std::make_pair(layerParams.name, out));
continue;
}
else if (layer_type == "Reshape")
{
CV_Assert(node_proto.input_size() == 2 || layerParams.has("shape"));
if (node_proto.input_size() == 2) {
Mat blob = getBlob(node_proto, constBlobs, 1);
CV_Assert(blob.type() == CV_32SC1);
if (layer_id.find(node_proto.input(0)) == layer_id.end()) {
Mat input = getBlob(node_proto, constBlobs, 0);
Mat out = input.reshape(0, static_cast<std::vector<int> >(blob));
constBlobs.insert(std::make_pair(layerParams.name, out));
continue;
}
layerParams.set("dim", DictValue::arrayInt<int*>(
blob.ptr<int>(), blob.total() ));
}
else {
DictValue shape = layerParams.get("shape");
std::vector<int> dim;
for (int j = 0; j < shape.size(); j++) {
dim.push_back(shape.getIntValue(j));
}
if (layer_id.find(node_proto.input(0)) == layer_id.end()) {
Mat input = getBlob(node_proto, constBlobs, 0);
Mat out = input.reshape(0, dim);
constBlobs.insert(std::make_pair(layerParams.name, out));
continue;
}
replaceLayerParam(layerParams, "shape", "dim");
}
}
else if (layer_type == "Pad")
{
layerParams.type = "Padding";
}
else if (layer_type == "Shape")
{
CV_Assert(node_proto.input_size() == 1);
shapeIt = outShapes.find(node_proto.input(0));
CV_Assert(shapeIt != outShapes.end());
MatShape inpShape = shapeIt->second;
Mat shapeMat(inpShape.size(), 1, CV_32S);
for (int j = 0; j < inpShape.size(); ++j)
shapeMat.at<int>(j) = inpShape[j];
shapeMat.dims = 1;
constBlobs.insert(std::make_pair(layerParams.name, shapeMat));
continue;
}
else if (layer_type == "Gather")
{
CV_Assert(node_proto.input_size() == 2);
CV_Assert(layerParams.has("axis"));
Mat input = getBlob(node_proto, constBlobs, 0);
Mat indexMat = getBlob(node_proto, constBlobs, 1);
CV_Assert_N(indexMat.type() == CV_32S, indexMat.total() == 1);
int index = indexMat.at<int>(0);
int axis = layerParams.get<int>("axis");
std::vector<cv::Range> ranges(input.dims, Range::all());
ranges[axis] = Range(index, index + 1);
Mat out = input(ranges);
constBlobs.insert(std::make_pair(layerParams.name, out));
continue;
}
else if (layer_type == "Concat")
{
bool hasVariableInps = false;
for (int i = 0; i < node_proto.input_size(); ++i)
{
if (layer_id.find(node_proto.input(i)) != layer_id.end())
{
hasVariableInps = true;
break;
}
}
if (!hasVariableInps)
{
std::vector<Mat> inputs(node_proto.input_size()), concatenated;
for (size_t i = 0; i < inputs.size(); ++i)
{
inputs[i] = getBlob(node_proto, constBlobs, i);
}
Ptr<Layer> concat = ConcatLayer::create(layerParams);
runLayer(concat, inputs, concatenated);
CV_Assert(concatenated.size() == 1);
constBlobs.insert(std::make_pair(layerParams.name, concatenated[0]));
continue;
}
}
else
{
for (int j = 0; j < node_proto.input_size(); j++) {
if (layer_id.find(node_proto.input(j)) == layer_id.end())
layerParams.blobs.push_back(getBlob(node_proto, constBlobs, j));
}
}
int id = dstNet.addLayer(layerParams.name, layerParams.type, layerParams);
layer_id.insert(std::make_pair(layerParams.name, LayerInfo(id, 0)));
std::vector<MatShape> layerInpShapes, layerOutShapes, layerInternalShapes;
for (int j = 0; j < node_proto.input_size(); j++) {
layerId = layer_id.find(node_proto.input(j));
if (layerId != layer_id.end()) {
dstNet.connect(layerId->second.layerId, layerId->second.outputId, id, j);
// Collect input shapes.
shapeIt = outShapes.find(node_proto.input(j));
CV_Assert(shapeIt != outShapes.end());
layerInpShapes.push_back(shapeIt->second);
}
}
// Compute shape of output blob for this layer.
Ptr<Layer> layer = dstNet.getLayer(id);
layer->getMemoryShapes(layerInpShapes, 0, layerOutShapes, layerInternalShapes);
CV_Assert(!layerOutShapes.empty());
outShapes[layerParams.name] = layerOutShapes[0];
}
}
