Fixing possible corruption for big training data sizes
This commit is contained in:
parent
e9539061db
commit
447b8bf58a
@ -777,7 +777,8 @@ cvGetTrainSamples( const CvMat* train_data, int tflag,
|
||||
|
||||
__BEGIN__;
|
||||
|
||||
int i, j, var_count, sample_count, s_step, v_step;
|
||||
int i, j, var_count, sample_count;
|
||||
size_t s_step, v_step, s;
|
||||
bool copy_data;
|
||||
const float* data;
|
||||
const int *s_idx, *v_idx;
|
||||
@ -815,7 +816,7 @@ cvGetTrainSamples( const CvMat* train_data, int tflag,
|
||||
{
|
||||
samples[0] = (float*)(samples + sample_count);
|
||||
if( tflag != CV_ROW_SAMPLE )
|
||||
CV_SWAP( s_step, v_step, i );
|
||||
CV_SWAP( s_step, v_step, s );
|
||||
|
||||
for( i = 0; i < sample_count; i++ )
|
||||
{
|
||||
|
Loading…
Reference in New Issue
Block a user