added new --discard-estimating-constant option, and a new car model

bin/dpmcreate.c

@@ -28,6 +28,7 @@ void exit_with_help()
 	"    --negative-cache-size : the cache size for negative examples it should be smaller than negative-count and larger than 100 [DEFAULT TO 2000]\n"
 	"    --include-overlap : the percentage of overlap between expected bounding box and the bounding box from detection. Beyond this threshold, it is ensured to be the same object [DEFAULT TO 0.7]\n"
 	"    --grayscale : 0 or 1, whether to exploit color in a given image [DEFAULT TO 0]\n"
+	"    --discard-estimating-constant : 0 or 1, when estimating bounding boxes, discarding constant (which may be accumulated error) [DEFAULT TO 1]\n"
 	"    --percentile-breakdown : 0.00 - 1.00, the percentile use for breakdown threshold [DEFAULT TO 0.05]\n\n"
 	);
 	exit(-1);
@@ -60,6 +61,7 @@ int main(int argc, char** argv)
 		{"percentile-breakdown", 1, 0, 0},
 		{"include-overlap", 1, 0, 0},
 		{"grayscale", 1, 0, 0},
+		{"discard-estimating-constant", 1, 0, 0},
 		{0, 0, 0, 0}
 	};
 	char* positive_list = 0;
@@ -68,24 +70,27 @@ int main(int argc, char** argv)
 	char* base_dir = 0;
 	int negative_count = 0;
 	ccv_dpm_param_t detector = { .interval = 8, .min_neighbors = 0, .flags = 0, .threshold = 0.0 };
-	ccv_dpm_new_param_t params = { .components = 0,
-								   .detector = detector,
-								   .parts = 0,
-								   .min_area = 3000,
-								   .max_area = 5000,
-								   .symmetric = 1,
-								   .alpha = 0.01,
-								   .balance = 1.5,
-								   .alpha_ratio = 0.995,
-								   .iterations = 1000,
-								   .data_minings = 50,
-								   .root_relabels = 20,
-								   .relabels = 10,
-								   .negative_cache_size = 2000,
-								   .C = 0.002,
-								   .percentile_breakdown = 0.05,
-								   .include_overlap = 0.7,
-								   .grayscale = 0 };
+	ccv_dpm_new_param_t params = {
+		.components = 0,
+		.detector = detector,
+		.parts = 0,
+		.min_area = 3000,
+		.max_area = 5000,
+		.symmetric = 1,
+		.alpha = 0.01,
+		.balance = 1.5,
+		.alpha_ratio = 0.995,
+		.iterations = 1000,
+		.data_minings = 50,
+		.root_relabels = 20,
+		.relabels = 10,
+		.negative_cache_size = 2000,
+		.C = 0.002,
+		.percentile_breakdown = 0.05,
+		.include_overlap = 0.7,
+		.grayscale = 0,
+		.discard_estimating_constant = 1,
+	};
 	int i, k;
 	while (getopt_long_only(argc, argv, "", dpm_options, &k) != -1)
 	{
@@ -152,6 +157,9 @@ int main(int argc, char** argv)
 			case 20:
 				params.grayscale = !!atoi(optarg);
 				break;
+			case 21:
+				params.discard_estimating_constant = !!atoi(optarg);
+				break;
 		}
 	}
 	assert(positive_list != 0);

bin/dpmdetect.c

@@ -21,7 +21,7 @@ int main(int argc, char** argv)
 		.interval = 8,
 		.min_neighbors = 1,
 		.flags = 0,
-		.threshold = 0.8
+		.threshold = 0.6
 	};
 	if (image != 0)
 	{

bin/dpmvldtr.rb

@@ -8,6 +8,7 @@
 File.new(ARGV[0]).each_line do |line|
 	args = line.split " "
 	name, x, y, width, height = args[0].to_s, args[1].to_i, args[2].to_i, args[3].to_i, args[4].to_i
+#	next if width < 44 or height < 28
 	truth[name] = Array.new unless truth.has_key? name
 	truth[name] << { :found => false, :x => x, :y => y, :width => width, :height => height }
 end

lib/ccv.h

@@ -710,6 +710,7 @@ typedef struct {
 	int data_minings;
 	int root_relabels;
 	int relabels;
+	int discard_estimating_constant; // 1
 	int negative_cache_size; // 1000
 	double include_overlap; // 0.7
 	double alpha;

lib/ccv_dpm.c

@@ -952,6 +952,8 @@ static ccv_array_t* _ccv_dpm_collect_all(gsl_rng* rng, ccv_dense_matrix_t* image
 						ccv_dpm_feature_vector_t* v = (ccv_dpm_feature_vector_t*)ccmalloc(sizeof(ccv_dpm_feature_vector_t));
 						_ccv_dpm_initialize_feature_vector_on_pattern(v, root_classifier, order[i]);
 						_ccv_dpm_collect_feature_vector(v, f_ptr[x] + root_classifier->beta, x, y, pyr[j], pyr[j - next], dx, dy);
+						v->scale_x = scale_x;
+						v->scale_y = scale_y;
 						ccv_array_push(av, &v);
 						if (av->rnum >= enough * (i + 1))
 							break;
@@ -1048,7 +1050,7 @@ static void _ccv_dpm_initialize_root_rectangle_estimator(ccv_dpm_mixture_model_t
 				ccv_rect_t bbox = bboxes[j];
 				gsl_vector_set(y[0], c, (bbox.x + bbox.width * 0.5) / (v->scale_x * CCV_DPM_WINDOW_SIZE) - v->x);
 				gsl_vector_set(y[1], c, (bbox.y + bbox.height * 0.5) / (v->scale_y * CCV_DPM_WINDOW_SIZE) - v->y);
-				gsl_vector_set(y[2], c, (bbox.width + bbox.height) / (root_classifier->root.w->rows * CCV_DPM_WINDOW_SIZE * v->scale_y + root_classifier->root.w->cols * CCV_DPM_WINDOW_SIZE * v->scale_x) - 1.0);
+				gsl_vector_set(y[2], c, sqrt((bbox.width * bbox.height) / (root_classifier->root.w->rows * v->scale_x * CCV_DPM_WINDOW_SIZE * root_classifier->root.w->cols * v->scale_y * CCV_DPM_WINDOW_SIZE)) - 1.0);
 				++c;
 			}
 		}
@@ -1057,7 +1059,7 @@ static void _ccv_dpm_initialize_root_rectangle_estimator(ccv_dpm_mixture_model_t
 		for (j = 0; j < 3; j++)
 		{
 			gsl_multifit_linear(X, y[j], z, cov, &chisq, workspace);
-			root_classifier->alpha[j] = gsl_vector_get(z, 0);
+			root_classifier->alpha[j] = params.discard_estimating_constant ? 0 : gsl_vector_get(z, 0);
 			for (k = 0; k < root_classifier->count; k++)
 			{
 				ccv_dpm_part_classifier_t* part_classifier = root_classifier->part + k;
@@ -2102,7 +2104,7 @@ ccv_array_t* ccv_dpm_detect_objects(ccv_dense_matrix_t* a, ccv_dpm_mixture_model
 								comp.part[k].rect = ccv_rect((int)((rx - pww) * CCV_DPM_WINDOW_SIZE / 2 * scale_x + 0.5), (int)((ry - pwh) * CCV_DPM_WINDOW_SIZE / 2 * scale_y + 0.5), (int)(part->w->cols * CCV_DPM_WINDOW_SIZE / 2 * scale_x + 0.5), (int)(part->w->rows * CCV_DPM_WINDOW_SIZE / 2 * scale_y + 0.5));
 								comp.part[k].confidence = -ccv_get_dense_matrix_cell_value_by(CCV_32F | CCV_C1, part_feature[k], iy, ix, 0);
 							}
-							comp.rect = ccv_rect((int)((x + drift_x) * CCV_DPM_WINDOW_SIZE * scale_x - rww * CCV_DPM_WINDOW_SIZE * (scale_x + drift_scale) + 0.5), (int)((y + drift_y) * CCV_DPM_WINDOW_SIZE * scale_y - rwh * CCV_DPM_WINDOW_SIZE * (scale_y + drift_scale) + 0.5), (int)(root->root.w->cols * CCV_DPM_WINDOW_SIZE * (scale_x + drift_scale) + 0.5), (int)(root->root.w->rows * CCV_DPM_WINDOW_SIZE * (scale_y + drift_scale) + 0.5));
+							comp.rect = ccv_rect((int)((x + drift_x) * CCV_DPM_WINDOW_SIZE * scale_x - rww * CCV_DPM_WINDOW_SIZE * scale_x * (1.0 + drift_scale) + 0.5), (int)((y + drift_y) * CCV_DPM_WINDOW_SIZE * scale_y - rwh * CCV_DPM_WINDOW_SIZE * scale_y * (1.0 + drift_scale) + 0.5), (int)(root->root.w->cols * CCV_DPM_WINDOW_SIZE * scale_x * (1.0 + drift_scale) + 0.5), (int)(root->root.w->rows * CCV_DPM_WINDOW_SIZE * scale_y * (1.0 + drift_scale) + 0.5));
 							ccv_array_push(seq, &comp);
 						}
 					f_ptr += root_feature->cols;