Merge pull request #687 from vpisarev:fast_lin_svm2

modules/ml/src/svm.cpp

@@ -1551,25 +1551,28 @@ void CvSVM::optimize_linear_svm()
         return;
 
     int var_count = get_var_count();
-    int sample_size = (int)(var_count*sizeof(sv[0][0]));
+    cv::AutoBuffer<double> vbuf(var_count);
+    double* v = vbuf;
     float** new_sv = (float**)cvMemStorageAlloc(storage, df_count*sizeof(new_sv[0]));
 
     for( i = 0; i < df_count; i++ )
     {
-        new_sv[i] = (float*)cvMemStorageAlloc(storage, sample_size);
+        new_sv[i] = (float*)cvMemStorageAlloc(storage, var_count*sizeof(new_sv[i][0]));
         float* dst = new_sv[i];
-        memset(dst, 0, sample_size);
+        memset(v, 0, var_count*sizeof(v[0]));
         int j, k, sv_count = df[i].sv_count;
         for( j = 0; j < sv_count; j++ )
         {
-            const float* src = class_count > 1 ? sv[df[i].sv_index[j]] : sv[j];
+            const float* src = class_count > 1 && df[i].sv_index ? sv[df[i].sv_index[j]] : sv[j];
             double a = df[i].alpha[j];
             for( k = 0; k < var_count; k++ )
-                dst[k] = (float)(dst[k] + src[k]*a);
+                v[k] += src[k]*a;
         }
+        for( k = 0; k < var_count; k++ )
+            dst[k] = (float)v[k];
         df[i].sv_count = 1;
         df[i].alpha[0] = 1.;
-        if( class_count > 1 )
+        if( class_count > 1 && df[i].sv_index )
             df[i].sv_index[0] = i;
     }
 
@@ -2570,6 +2573,7 @@ void CvSVM::read( CvFileStorage* fs, CvFileNode* svm_node )
         CV_NEXT_SEQ_ELEM( df_node->data.seq->elem_size, reader );
     }
 
+    if( cvReadIntByName(fs, svm_node, "optimize_linear", 1) != 0 )
         optimize_linear_svm();
     create_kernel();
 

modules/ml/test/test_mltests2.cpp

@@ -769,7 +769,11 @@ void CV_MLBaseTest::load( const char* filename )
     else if( !modelName.compare(CV_KNEAREST) )
         knearest->load( filename );
     else if( !modelName.compare(CV_SVM) )
+    {
+        delete svm;
+        svm = new CvSVM;
         svm->load( filename );
+    }
     else if( !modelName.compare(CV_ANN) )
         ann->load( filename );
     else if( !modelName.compare(CV_DTREE) )

modules/ml/test/test_save_load.cpp

@@ -82,32 +82,53 @@ int CV_SLMLTest::validate_test_results( int testCaseIdx )
     int code = cvtest::TS::OK;
 
     // 1. compare files
-    ifstream f1( fname1.c_str() ), f2( fname2.c_str() );
+    FILE *fs1 = fopen(fname1.c_str(), "rb"), *fs2 = fopen(fname2.c_str(), "rb");
-    string s1, s2;
+    size_t sz1 = 0, sz2 = 0;
-    int lineIdx = 0;
+    if( !fs1 || !fs2 )
-    CV_Assert( f1.is_open() && f2.is_open() );
+        code = cvtest::TS::FAIL_MISSING_TEST_DATA;
-    for( ; !f1.eof() && !f2.eof(); lineIdx++ )
+    if( code >= 0 )
     {
-        getline( f1, s1 );
+        fseek(fs1, 0, SEEK_END); fseek(fs2, 0, SEEK_END);
-        getline( f2, s2 );
+        sz1 = ftell(fs1);
-        if( s1.compare(s2) )
+        sz2 = ftell(fs2);
+        fseek(fs1, 0, SEEK_SET); fseek(fs2, 0, SEEK_SET);
+    }
+
+    if( sz1 != sz2 )
+        code = cvtest::TS::FAIL_INVALID_OUTPUT;
+
+    if( code >= 0 )
+    {
+        const int BUFSZ = 1024;
+        uchar buf1[BUFSZ], buf2[BUFSZ];
+        for( size_t pos = 0; pos < sz1;  )
+        {
+            size_t r1 = fread(buf1, 1, BUFSZ, fs1);
+            size_t r2 = fread(buf2, 1, BUFSZ, fs2);
+            if( r1 != r2 || memcmp(buf1, buf2, r1) != 0 )
             {
-            ts->printf( cvtest::TS::LOG, "first and second saved files differ in %n-line; first %n line: %s; second %n-line: %s",
+                ts->printf( cvtest::TS::LOG,
-               lineIdx, lineIdx, s1.c_str(), lineIdx, s2.c_str() );
+                           "in test case %d first (%s) and second (%s) saved files differ in %d-th kb\n",
+                           testCaseIdx, fname1.c_str(), fname2.c_str(),
+                           (int)pos );
                 code = cvtest::TS::FAIL_INVALID_OUTPUT;
+                break;
             }
+            pos += r1;
         }
-    if( !f1.eof() || !f2.eof() )
-    {
-        ts->printf( cvtest::TS::LOG, "in test case %d first and second saved files differ in %n-line; first %n line: %s; second %n-line: %s",
-            testCaseIdx, lineIdx, lineIdx, s1.c_str(), lineIdx, s2.c_str() );
-        code = cvtest::TS::FAIL_INVALID_OUTPUT;
     }
-    f1.close();
+
-    f2.close();
+    if(fs1)
+        fclose(fs1);
+    if(fs2)
+        fclose(fs2);
+
     // delete temporary files
+    if( code >= 0 )
+    {
         remove( fname1.c_str() );
         remove( fname2.c_str() );
+    }
 
     // 2. compare responses
     CV_Assert( test_resps1.size() == test_resps2.size() );
@@ -133,4 +154,32 @@ TEST(ML_Boost, save_load) { CV_SLMLTest test( CV_BOOST ); test.safe_run(); }
 TEST(ML_RTrees, save_load) { CV_SLMLTest test( CV_RTREES ); test.safe_run(); }
 TEST(ML_ERTrees, save_load) { CV_SLMLTest test( CV_ERTREES ); test.safe_run(); }
 
+
+TEST(DISABLED_ML_SVM, linear_save_load)
+{
+    CvSVM svm1, svm2, svm3;
+    svm1.load("SVM45_X_38-1.xml");
+    svm2.load("SVM45_X_38-2.xml");
+    string tname = tempfile("a.xml");
+    svm2.save(tname.c_str());
+    svm3.load(tname.c_str());
+
+    ASSERT_EQ(svm1.get_var_count(), svm2.get_var_count());
+    ASSERT_EQ(svm1.get_var_count(), svm3.get_var_count());
+
+    int m = 10000, n = svm1.get_var_count();
+    Mat samples(m, n, CV_32F), r1, r2, r3;
+    randu(samples, 0., 1.);
+
+    svm1.predict(samples, r1);
+    svm2.predict(samples, r2);
+    svm3.predict(samples, r3);
+
+    double eps = 1e-4;
+    EXPECT_LE(norm(r1, r2, NORM_INF), eps);
+    EXPECT_LE(norm(r1, r3, NORM_INF), eps);
+
+    remove(tname.c_str());
+}
+
 /* End of file. */