END_TEST
START_TEST (document_type)
{
reset_errno();
size_t c = model_size(model);
assert_noerr();
assert_uint_eq(1, c);
reset_errno();
Document *d = model_document(model, 0);
assert_noerr();
assert_not_null(d);
reset_errno();
Node *bogus = model_document(model, 1);
assert_errno(EINVAL);
assert_null(bogus);
reset_errno();
Node *r1 = model_document_root(model, 0);
assert_noerr();
assert_not_null(r1);
assert_node_kind(r1, MAPPING);
reset_errno();
Node *r2 = document_root(d);
assert_noerr();
assert_not_null(r2);
assert_ptr_eq(r1, r2);
}
void ICFDetector::train(const vector<string>& image_filenames,
const vector< vector<Rect> >& labelling,
ICFDetectorParams params)
{
Size model_size(params.model_n_cols, params.model_n_rows);
vector<Mat> samples; /* positive samples + negative samples */
Mat sample, resized_sample;
int pos_count = 0;
for( size_t i = 0; i < image_filenames.size(); ++i, ++pos_count )
{
Mat img = imread(String(image_filenames[i].c_str()));
for( size_t j = 0; j < labelling[i].size(); ++j )
{
Rect r = labelling[i][j];
if( r.x > img.cols || r.y > img.rows )
continue;
sample = img.colRange(max(r.x, 0), min(r.width, img.cols))
.rowRange(max(r.y, 0), min(r.height, img.rows));
resize(sample, resized_sample, model_size);
samples.push_back(resized_sample);
}
}
int neg_count = 0;
RNG rng;
for( size_t i = 0; i < image_filenames.size(); ++i )
{
Mat img = imread(String(image_filenames[i].c_str()));
for( int j = 0; j < (int)(pos_count / image_filenames.size() + 1); )
{
Rect r;
r.x = rng.uniform(0, img.cols);
r.width = rng.uniform(r.x + 1, img.cols);
r.y = rng.uniform(0, img.rows);
r.height = rng.uniform(r.y + 1, img.rows);
if( !overlap(r, labelling[i]) )
{
sample = img.colRange(r.x, r.width).rowRange(r.y, r.height);
//resize(sample, resized_sample);
samples.push_back(resized_sample);
++neg_count;
++j;
}
}
}
Mat_<int> labels(1, pos_count + neg_count);
for( int i = 0; i < pos_count; ++i)
labels(0, i) = 1;
for( int i = pos_count; i < pos_count + neg_count; ++i )
labels(0, i) = -1;
vector<Point3i> features = generateFeatures(model_size);
Ptr<ACFFeatureEvaluator> feature_evaluator = createACFFeatureEvaluator(features);
Mat_<int> data((int)features.size(), (int)samples.size());
Mat_<int> feature_col;
vector<Mat> channels;
for( int i = 0; i < (int)samples.size(); ++i )
{
computeChannels(samples[i], channels);
feature_evaluator->setChannels(channels);
feature_evaluator->evaluateAll(feature_col);
for( int j = 0; j < feature_col.rows; ++j )
data(i, j) = feature_col(0, j);
}
WaldBoostParams wparams;
wparams.weak_count = params.weak_count;
wparams.alpha = 0.001f;
Ptr<WaldBoost> waldboost = createWaldBoost(wparams);
waldboost->train(data, labels);
}
