Files
opencv-MIRROR/modules/ml/test/test_svmtrainauto.cpp
Yixuan Tang ce4ceac609 ml(svm): fix UB pointer overflow in solve_generic with extreme float data (#29111)
* ml: guard against negative indices in SVM solve_generic

* ml: reject non-finite SVM solver state and invalid working sets

---------

Co-authored-by: example <example@example.com>
2026-11-13 10:47:37 +05:30

303 lines
9.8 KiB
C++

// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
#include "test_precomp.hpp"
#include <limits>
namespace opencv_test { namespace {
using cv::ml::SVM;
using cv::ml::TrainData;
static Ptr<TrainData> makeRandomData(int datasize)
{
cv::Mat samples = cv::Mat::zeros( datasize, 2, CV_32FC1 );
cv::Mat responses = cv::Mat::zeros( datasize, 1, CV_32S );
RNG &rng = cv::theRNG();
for (int i = 0; i < datasize; ++i)
{
int response = rng.uniform(0, 2); // Random from {0, 1}.
samples.at<float>( i, 0 ) = rng.uniform(0.f, 0.5f) + response * 0.5f;
samples.at<float>( i, 1 ) = rng.uniform(0.f, 0.5f) + response * 0.5f;
responses.at<int>( i, 0 ) = response;
}
return TrainData::create( samples, cv::ml::ROW_SAMPLE, responses );
}
static Ptr<TrainData> makeCircleData(int datasize, float scale_factor, float radius)
{
// Populate samples with data that can be split into two concentric circles
cv::Mat samples = cv::Mat::zeros( datasize, 2, CV_32FC1 );
cv::Mat responses = cv::Mat::zeros( datasize, 1, CV_32S );
for (int i = 0; i < datasize; i+=2)
{
const float pi = 3.14159f;
const float angle_rads = (i/datasize) * pi;
const float x = radius * cos(angle_rads);
const float y = radius * cos(angle_rads);
// Larger circle
samples.at<float>( i, 0 ) = x;
samples.at<float>( i, 1 ) = y;
responses.at<int>( i, 0 ) = 0;
// Smaller circle
samples.at<float>( i + 1, 0 ) = x * scale_factor;
samples.at<float>( i + 1, 1 ) = y * scale_factor;
responses.at<int>( i + 1, 0 ) = 1;
}
return TrainData::create( samples, cv::ml::ROW_SAMPLE, responses );
}
static Ptr<TrainData> makeRandomData2(int datasize)
{
cv::Mat samples = cv::Mat::zeros( datasize, 2, CV_32FC1 );
cv::Mat responses = cv::Mat::zeros( datasize, 1, CV_32S );
RNG &rng = cv::theRNG();
for (int i = 0; i < datasize; ++i)
{
int response = rng.uniform(0, 2); // Random from {0, 1}.
samples.at<float>( i, 0 ) = 0;
samples.at<float>( i, 1 ) = (0.5f - response) * rng.uniform(0.f, 1.2f) + response;
responses.at<int>( i, 0 ) = response;
}
return TrainData::create( samples, cv::ml::ROW_SAMPLE, responses );
}
//==================================================================================================
TEST(ML_SVM, trainauto)
{
const int datasize = 100;
cv::Ptr<TrainData> data = makeRandomData(datasize);
ASSERT_TRUE(data);
cv::Ptr<SVM> svm = SVM::create();
ASSERT_TRUE(svm);
svm->trainAuto( data, 10 ); // 2-fold cross validation.
float test_data0[2] = {0.25f, 0.25f};
cv::Mat test_point0 = cv::Mat( 1, 2, CV_32FC1, test_data0 );
float result0 = svm->predict( test_point0 );
float test_data1[2] = {0.75f, 0.75f};
cv::Mat test_point1 = cv::Mat( 1, 2, CV_32FC1, test_data1 );
float result1 = svm->predict( test_point1 );
EXPECT_NEAR(result0, 0, 0.001);
EXPECT_NEAR(result1, 1, 0.001);
}
TEST(ML_SVM, trainauto_sigmoid)
{
const int datasize = 100;
const float scale_factor = 0.5;
const float radius = 2.0;
cv::Ptr<TrainData> data = makeCircleData(datasize, scale_factor, radius);
ASSERT_TRUE(data);
cv::Ptr<SVM> svm = SVM::create();
ASSERT_TRUE(svm);
svm->setKernel(SVM::SIGMOID);
svm->setGamma(10.0);
svm->setCoef0(-10.0);
svm->trainAuto( data, 10 ); // 2-fold cross validation.
float test_data0[2] = {radius, radius};
cv::Mat test_point0 = cv::Mat( 1, 2, CV_32FC1, test_data0 );
EXPECT_FLOAT_EQ(svm->predict( test_point0 ), 0);
float test_data1[2] = {scale_factor * radius, scale_factor * radius};
cv::Mat test_point1 = cv::Mat( 1, 2, CV_32FC1, test_data1 );
EXPECT_FLOAT_EQ(svm->predict( test_point1 ), 1);
}
TEST(ML_SVM, trainAuto_regression_5369)
{
const int datasize = 100;
Ptr<TrainData> data = makeRandomData2(datasize);
cv::Ptr<SVM> svm = SVM::create();
svm->trainAuto( data, 10 ); // 2-fold cross validation.
float test_data0[2] = {0.25f, 0.25f};
cv::Mat test_point0 = cv::Mat( 1, 2, CV_32FC1, test_data0 );
float result0 = svm->predict( test_point0 );
float test_data1[2] = {0.75f, 0.75f};
cv::Mat test_point1 = cv::Mat( 1, 2, CV_32FC1, test_data1 );
float result1 = svm->predict( test_point1 );
EXPECT_EQ(0., result0);
EXPECT_EQ(1., result1);
}
TEST(ML_SVM, getSupportVectors)
{
// Set up training data
int labels[4] = {1, -1, -1, -1};
float trainingData[4][2] = { {501, 10}, {255, 10}, {501, 255}, {10, 501} };
Mat trainingDataMat(4, 2, CV_32FC1, trainingData);
Mat labelsMat(4, 1, CV_32SC1, labels);
Ptr<SVM> svm = SVM::create();
ASSERT_TRUE(svm);
svm->setType(SVM::C_SVC);
svm->setTermCriteria(TermCriteria(TermCriteria::MAX_ITER, 100, 1e-6));
// Test retrieval of SVs and compressed SVs on linear SVM
svm->setKernel(SVM::LINEAR);
svm->train(trainingDataMat, cv::ml::ROW_SAMPLE, labelsMat);
Mat sv = svm->getSupportVectors();
EXPECT_EQ(1, sv.rows); // by default compressed SV returned
sv = svm->getUncompressedSupportVectors();
EXPECT_EQ(3, sv.rows);
// Test retrieval of SVs and compressed SVs on non-linear SVM
svm->setKernel(SVM::POLY);
svm->setDegree(2);
svm->train(trainingDataMat, cv::ml::ROW_SAMPLE, labelsMat);
sv = svm->getSupportVectors();
EXPECT_EQ(3, sv.rows);
sv = svm->getUncompressedSupportVectors();
EXPECT_EQ(0, sv.rows); // inapplicable for non-linear SVMs
}
TEST(ML_SVM, nu_svc_extreme_float_no_crash)
{
float fdata[] = {
-3.40282347e+38f, 0.0f,
3.08375703e+38f, 0.0f,
3.38958311e+38f, 0.0f,
-3.40053886e+38f, 0.0f,
-5.19229686e+33f, 0.0f,
-5.17201445e+33f, 0.0f,
};
int labels[] = {1, 0, 0, 0, 0, 0};
Mat features(6, 2, CV_32FC1, fdata);
Mat responses(6, 1, CV_32SC1, labels);
Ptr<SVM> svm = SVM::create();
ASSERT_TRUE(svm);
svm->setType(SVM::NU_SVC);
svm->setKernel(SVM::LINEAR);
svm->setNu(0.010317);
svm->setTermCriteria(TermCriteria(
TermCriteria::MAX_ITER | TermCriteria::EPS, 11, 0.099998f));
EXPECT_FALSE(svm->train(features, cv::ml::ROW_SAMPLE, responses));
EXPECT_FALSE(svm->isTrained());
EXPECT_TRUE(svm->getSupportVectors().empty());
}
class NonFiniteSVMKernel : public SVM::Kernel
{
public:
explicit NonFiniteSVMKernel(float value) : value_(value), calls(0) {}
int getType() const CV_OVERRIDE { return SVM::CUSTOM; }
void calc(int vcount, int, const float*, const float*, float* results) CV_OVERRIDE
{
++calls;
for (int i = 0; i < vcount; ++i)
results[i] = value_;
}
float value_;
int calls;
};
TEST(ML_SVM, nu_svc_non_finite_gradient)
{
// The built-in kernels clamp NaNs; use a custom kernel to reach the solver checks.
const float values[] = {
std::numeric_limits<float>::quiet_NaN(),
std::numeric_limits<float>::infinity(),
-std::numeric_limits<float>::infinity()
};
Mat features = (Mat_<float>(4, 2) << -2, -1, -1, -2, 1, 2, 2, 1);
Mat responses = (Mat_<int>(4, 1) << 0, 0, 1, 1);
for (size_t i = 0; i < sizeof(values) / sizeof(values[0]); ++i)
{
SCOPED_TRACE(i);
Ptr<NonFiniteSVMKernel> kernel = makePtr<NonFiniteSVMKernel>(values[i]);
Ptr<SVM> svm = SVM::create();
svm->setType(SVM::NU_SVC);
svm->setNu(0.5);
svm->setCustomKernel(kernel);
EXPECT_FALSE(svm->train(features, cv::ml::ROW_SAMPLE, responses));
EXPECT_GT(kernel->calls, 0);
EXPECT_FALSE(svm->isTrained());
EXPECT_TRUE(svm->getSupportVectors().empty());
}
}
TEST(ML_SVM, finite_linear_training)
{
Mat features = (Mat_<float>(4, 2) << -2, -1, -1, -2, 1, 2, 2, 1);
Mat responses = (Mat_<int>(4, 1) << 0, 0, 1, 1);
Mat queries = (Mat_<float>(2, 2) << -3, -3, 3, 3);
const int types[] = { SVM::C_SVC, SVM::NU_SVC };
for (size_t i = 0; i < sizeof(types) / sizeof(types[0]); ++i)
{
SCOPED_TRACE(types[i]);
Ptr<SVM> svm = SVM::create();
svm->setType(types[i]);
svm->setKernel(SVM::LINEAR);
svm->setNu(0.5);
ASSERT_TRUE(svm->train(features, cv::ml::ROW_SAMPLE, responses));
EXPECT_TRUE(svm->isTrained());
EXPECT_FALSE(svm->getSupportVectors().empty());
EXPECT_FLOAT_EQ(0, svm->predict(queries.row(0)));
EXPECT_FLOAT_EQ(1, svm->predict(queries.row(1)));
}
}
TEST(ML_SVM, finite_linear_regression_training)
{
Mat features = (Mat_<float>(4, 1) << -2, -1, 1, 2);
Mat responses = (Mat_<float>(4, 1) << -3, -1, 3, 5);
Mat query = (Mat_<float>(1, 1) << 0.5f);
const int types[] = { SVM::EPS_SVR, SVM::NU_SVR };
for (size_t i = 0; i < sizeof(types) / sizeof(types[0]); ++i)
{
SCOPED_TRACE(types[i]);
Ptr<SVM> svm = SVM::create();
svm->setType(types[i]);
svm->setKernel(SVM::LINEAR);
svm->setC(10);
svm->setP(0.01);
svm->setNu(0.5);
ASSERT_TRUE(svm->train(features, cv::ml::ROW_SAMPLE, responses));
EXPECT_TRUE(svm->isTrained());
EXPECT_NEAR(2, svm->predict(query), 0.05);
}
}
TEST(ML_SVM, finite_linear_one_class_training)
{
Mat features = (Mat_<float>(4, 1) << 1, 2, 3, 4);
Mat responses = Mat::ones(4, 1, CV_32S);
Mat inlier = (Mat_<float>(1, 1) << 5);
Mat outlier = (Mat_<float>(1, 1) << -1);
Ptr<SVM> svm = SVM::create();
svm->setType(SVM::ONE_CLASS);
svm->setKernel(SVM::LINEAR);
svm->setNu(0.5);
ASSERT_TRUE(svm->train(features, cv::ml::ROW_SAMPLE, responses));
EXPECT_TRUE(svm->isTrained());
EXPECT_FLOAT_EQ(1, svm->predict(inlier));
EXPECT_FLOAT_EQ(0, svm->predict(outlier));
}
}} // namespace