Made changes to allow ml module to work with big data.

modules/ml/include/opencv2/ml/ml.hpp

@@ -796,7 +796,7 @@ struct CV_EXPORTS CvDTreeTrainData
     const CvMat* responses;
     CvMat* responses_copy; // used in Boosting
 
-    int buf_count, buf_size;
+    int buf_count, buf_size; // buf_size is obsolete, please do not use it, use expression ((int64)buf->rows * (int64)buf->cols / buf_count) instead
     bool shared;
     int is_buf_16u;
 
@@ -806,6 +806,12 @@ struct CV_EXPORTS CvDTreeTrainData
 
     CvMat* counts;
     CvMat* buf;
+    inline size_t get_length_subbuf() const
+    {
+        size_t res = (size_t)(work_var_count + 1) * (size_t)sample_count;
+        return res;
+    }
+
     CvMat* direction;
     CvMat* split_buf;
 

modules/ml/src/boost.cpp

@@ -1130,13 +1130,13 @@ CvBoost::update_weights( CvBoostTree* tree )
     int *sample_idx_buf;
     const int* sample_idx = 0;
     cv::AutoBuffer<uchar> inn_buf;
-    size_t _buf_size = (params.boost_type == LOGIT) || (params.boost_type == GENTLE) ? data->sample_count*sizeof(int) : 0;
+    size_t _buf_size = (params.boost_type == LOGIT) || (params.boost_type == GENTLE) ? (size_t)(data->sample_count)*sizeof(int) : 0;
     if( !tree )
         _buf_size += n*sizeof(int);
     else
     {
         if( have_subsample )
-            _buf_size += data->buf->cols*(sizeof(float)+sizeof(uchar));
+            _buf_size += data->get_length_subbuf()*(sizeof(float)+sizeof(uchar));
     }
     inn_buf.allocate(_buf_size);
     uchar* cur_buf_pos = (uchar*)inn_buf;
@@ -1151,6 +1151,7 @@ CvBoost::update_weights( CvBoostTree* tree )
         sample_idx = data->get_sample_indices( data->data_root, sample_idx_buf );
     }
     CvMat* dtree_data_buf = data->buf;
+    size_t length_buf_row = data->get_length_subbuf();
     if( !tree ) // before training the first tree, initialize weights and other parameters
     {
         int* class_labels_buf = (int*)cur_buf_pos;
@@ -1189,7 +1190,7 @@ CvBoost::update_weights( CvBoostTree* tree )
 
         if (data->is_buf_16u)
         {
-            unsigned short* labels = (unsigned short*)(dtree_data_buf->data.s + data->data_root->buf_idx*dtree_data_buf->cols +
+            unsigned short* labels = (unsigned short*)(dtree_data_buf->data.s + data->data_root->buf_idx*length_buf_row +
                 data->data_root->offset + (data->work_var_count-1)*data->sample_count);
             for( i = 0; i < n; i++ )
             {
@@ -1207,7 +1208,7 @@ CvBoost::update_weights( CvBoostTree* tree )
         }
         else
         {
-            int* labels = dtree_data_buf->data.i + data->data_root->buf_idx*dtree_data_buf->cols +
+            int* labels = dtree_data_buf->data.i + data->data_root->buf_idx*length_buf_row +
                 data->data_root->offset + (data->work_var_count-1)*data->sample_count;
 
             for( i = 0; i < n; i++ )
@@ -1254,9 +1255,10 @@ CvBoost::update_weights( CvBoostTree* tree )
         if( have_subsample )
         {
             float* values = (float*)cur_buf_pos;
-            cur_buf_pos = (uchar*)(values + data->buf->cols);
+            cur_buf_pos = (uchar*)(values + data->get_length_subbuf());
             uchar* missing = cur_buf_pos;
-            cur_buf_pos = missing + data->buf->step;
+            cur_buf_pos = missing + data->get_length_subbuf() * (size_t)CV_ELEM_SIZE(data->buf->type);
+
             CvMat _sample, _mask;
 
             // invert the subsample mask

modules/ml/src/ertrees.cpp

@@ -75,11 +75,14 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
     int sample_all = 0, r_type, cv_n;
     int total_c_count = 0;
     int tree_block_size, temp_block_size, max_split_size, nv_size, cv_size = 0;
-    int ds_step, dv_step, ms_step = 0, mv_step = 0; // {data|mask}{sample|var}_step
-    int vi, i, size;
+    int64 ds_step, dv_step, ms_step = 0, mv_step = 0; // {data|mask}{sample|var}_step
+    int64 vi, i, size;
     char err[100];
     const int *sidx = 0, *vidx = 0;
 
+    uint64 effective_buf_size = -1;
+    int effective_buf_height = -1, effective_buf_width = -1;
+
     if ( _params.use_surrogates )
         CV_ERROR(CV_StsBadArg, "CvERTrees do not support surrogate splits");
 
@@ -179,18 +182,34 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
     have_labels = cv_n > 0 || (ord_var_count == 1 && cat_var_count == 0) || _add_labels;
 
     work_var_count = cat_var_count + (is_classifier ? 1 : 0) + (have_labels ? 1 : 0);
-    buf_size = (work_var_count + 1)*sample_count;
+
     shared = _shared;
     buf_count = shared ? 2 : 1;
 
+    buf_size = -1; // the member buf_size is obsolete
+
+    effective_buf_size = (uint64)(work_var_count + 1)*(uint64)sample_count * buf_count; // this is the total size of "CvMat buf" to be allocated
+    effective_buf_width = sample_count;
+    effective_buf_height = work_var_count+1;
+
+    if (effective_buf_width >= effective_buf_height)
+        effective_buf_height *= buf_count;
+    else
+        effective_buf_width *= buf_count;
+
+    if ((uint64)effective_buf_width * (uint64)effective_buf_height != effective_buf_size)
+    {
+        CV_Error(CV_StsBadArg, "The memory buffer cannot be allocated since its size exceeds integer fields limit");
+    }
+
     if ( is_buf_16u )
     {
-        CV_CALL( buf = cvCreateMat( buf_count, buf_size, CV_16UC1 ));
+        CV_CALL( buf = cvCreateMat( effective_buf_height, effective_buf_width, CV_16UC1 ));
         CV_CALL( pair16u32s_ptr = (CvPair16u32s*)cvAlloc( sample_count*sizeof(pair16u32s_ptr[0]) ));
     }
     else
     {
-        CV_CALL( buf = cvCreateMat( buf_count, buf_size, CV_32SC1 ));
+        CV_CALL( buf = cvCreateMat( effective_buf_height, effective_buf_width, CV_32SC1 ));
         CV_CALL( int_ptr = (int**)cvAlloc( sample_count*sizeof(int_ptr[0]) ));
     }
 
@@ -303,7 +322,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
                         val = cvRound(t);
                         if( val != t )
                         {
-                            sprintf( err, "%d-th value of %d-th (categorical) "
+                            sprintf( err, "%ld-th value of %ld-th (categorical) "
                                 "variable is not an integer", i, vi );
                             CV_ERROR( CV_StsBadArg, err );
                         }
@@ -311,7 +330,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
 
                     if( val == INT_MAX )
                     {
-                        sprintf( err, "%d-th value of %d-th (categorical) "
+                        sprintf( err, "%ld-th value of %ld-th (categorical) "
                             "variable is too large", i, vi );
                         CV_ERROR( CV_StsBadArg, err );
                     }
@@ -414,7 +433,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
 
                     if( fabs(val) >= ord_nan )
                     {
-                        sprintf( err, "%d-th value of %d-th (ordered) "
+                        sprintf( err, "%ld-th value of %ld-th (ordered) "
                             "variable (=%g) is too large", i, vi, val );
                         CV_ERROR( CV_StsBadArg, err );
                     }
@@ -578,9 +597,9 @@ const int* CvERTreeTrainData::get_cat_var_data( CvDTreeNode* n, int vi, int* cat
     int ci = get_var_type( vi);
     const int* cat_values = 0;
     if( !is_buf_16u )
-        cat_values = buf->data.i + n->buf_idx*buf->cols + ci*sample_count + n->offset;
+        cat_values = buf->data.i + n->buf_idx*get_length_subbuf() + ci*sample_count + n->offset;
     else {
-        const unsigned short* short_values = (const unsigned short*)(buf->data.s + n->buf_idx*buf->cols +
+        const unsigned short* short_values = (const unsigned short*)(buf->data.s + n->buf_idx*get_length_subbuf() +
             ci*sample_count + n->offset);
         for( int i = 0; i < n->sample_count; i++ )
             cat_values_buf[i] = short_values[i];
@@ -1333,6 +1352,7 @@ void CvForestERTree::split_node_data( CvDTreeNode* node )
     CvDTreeNode *left = 0, *right = 0;
     int new_buf_idx = data->get_child_buf_idx( node );
     CvMat* buf = data->buf;
+    size_t length_buf_row = data->get_length_subbuf();
     cv::AutoBuffer<int> temp_buf(n);
 
     complete_node_dir(node);
@@ -1385,9 +1405,9 @@ void CvForestERTree::split_node_data( CvDTreeNode* node )
 
         if (data->is_buf_16u)
         {
-            unsigned short *ldst = (unsigned short *)(buf->data.s + left->buf_idx*buf->cols +
+            unsigned short *ldst = (unsigned short *)(buf->data.s + left->buf_idx*length_buf_row +
                 ci*scount + left->offset);
-            unsigned short *rdst = (unsigned short *)(buf->data.s + right->buf_idx*buf->cols +
+            unsigned short *rdst = (unsigned short *)(buf->data.s + right->buf_idx*length_buf_row +
                 ci*scount + right->offset);
 
             for( i = 0; i < n; i++ )
@@ -1415,9 +1435,9 @@ void CvForestERTree::split_node_data( CvDTreeNode* node )
         }
         else
         {
-            int *ldst = buf->data.i + left->buf_idx*buf->cols +
+            int *ldst = buf->data.i + left->buf_idx*length_buf_row +
                 ci*scount + left->offset;
-            int *rdst = buf->data.i + right->buf_idx*buf->cols +
+            int *rdst = buf->data.i + right->buf_idx*length_buf_row +
                 ci*scount + right->offset;
 
             for( i = 0; i < n; i++ )
@@ -1460,9 +1480,9 @@ void CvForestERTree::split_node_data( CvDTreeNode* node )
 
         if (data->is_buf_16u)
         {
-            unsigned short* ldst = (unsigned short*)(buf->data.s + left->buf_idx*buf->cols +
+            unsigned short* ldst = (unsigned short*)(buf->data.s + left->buf_idx*length_buf_row +
                 pos*scount + left->offset);
-            unsigned short* rdst = (unsigned short*)(buf->data.s + right->buf_idx*buf->cols +
+            unsigned short* rdst = (unsigned short*)(buf->data.s + right->buf_idx*length_buf_row +
                 pos*scount + right->offset);
 
             for (i = 0; i < n; i++)
@@ -1483,9 +1503,9 @@ void CvForestERTree::split_node_data( CvDTreeNode* node )
         }
         else
         {
-            int* ldst = buf->data.i + left->buf_idx*buf->cols +
+            int* ldst = buf->data.i + left->buf_idx*length_buf_row +
                 pos*scount + left->offset;
-            int* rdst = buf->data.i + right->buf_idx*buf->cols +
+            int* rdst = buf->data.i + right->buf_idx*length_buf_row +
                 pos*scount + right->offset;
             for (i = 0; i < n; i++)
             {