fullseye

HALCON operator coverage (measured vs the real reference)

Source: https://www.mvtec.com/doc/halcon/2605/en/ (version 2605). Ground truth: 2313 operators across 30 top-level chapters (218 TOC pages), mined by halcon_scrape.py.

imgevolve maps to 981 / 2313 HALCON operators (42.4%) via Op.halcon, from 860 registry ops.

One imgevolve op claims one nearest HALCON operator, so coverage counts distinct real operators with an analogue. This number is grounded in the scraped reference, not memory; grow it by adding operator families to the registry (each new Op.halcon that names a real operator lifts coverage).

Per-chapter coverage (ranked by gap)

chapter	covered	total	gap
Graphics	9	174	165
Tuple	0	165	165
System	0	141	141
Classification	0	99	99
OCR	0	96	96
Legacy	16	110	94
Deep Learning	4	88	84
Matching	38	96	58
File	0	53	53
3D Matching	15	59	44
Inspection	12	55	43
Develop	0	37	37
Control	0	34	34
Calibration	35	68	33
3D Reconstruction	50	76	26
Identification	1	27	26
Image Source	0	25	25
Image	88	110	22
2D Metrology	10	32	22
Tools	90	108	18
Object	1	16	15
Transformations	104	118	14
Matrix	46	57	11
3D Object Model	40	51	11
Filters	186	196	10
XLD	88	97	9
1D Measuring	11	20	9
Segmentation	48	53	5
Morphology	42	44	2
Regions	105	106	1

Build targets — biggest gaps first (sample uncovered operators)

• Graphics (9/174): add_scene_3d_camera, add_scene_3d_instance, add_scene_3d_label, add_scene_3d_light, attach_background_to_window, attach_drawing_object_to_window, clear_drawing_object, clear_scene_3d, clear_window, close_window, convert_coordinates_image_to_window, convert_coordinates_window_to_image
• Tuple (0/165): clear_handle, copy_dict, create_dict, dict_to_json, get_dict_object, get_dict_param, get_dict_tuple, get_handle_object, get_handle_param, get_handle_tuple, handle_to_integer, integer_to_handle
• System (0/141): activate_compute_device, broadcast_condition, clear_barrier, clear_condition, clear_event, clear_message, clear_message_queue, clear_mutex, clear_serial, clear_serialized_item, close_io_channel, close_io_device
• Classification (0/99): add_class_train_data_gmm, add_class_train_data_knn, add_class_train_data_mlp, add_class_train_data_svm, add_sample_class_gmm, add_sample_class_knn, add_sample_class_mlp, add_sample_class_svm, add_sample_class_train_data, classify_class_gmm, classify_class_knn, classify_class_mlp
• OCR (0/96): append_ocr_trainf, apply_deep_ocr, clear_lexicon, clear_ocr_class_cnn, clear_ocr_class_knn, clear_ocr_class_mlp, clear_ocr_class_svm, clear_text_model, clear_text_result, concat_ocr_trainf, create_deep_ocr, create_lexicon
• Legacy (16/110): approx_chain, approx_chain_simple, clear_component_model, clear_rectangle, clear_training_components, close_ocr, cluster_model_components, copy_metrology_object, create_component_model, create_ocr_class_box, create_text_model, create_trained_component_model
• Deep Learning (4/88): add_dl_pruning_batch, apply_dl_model, clear_dl_model, create_dl_layer_activation, create_dl_layer_affine_grid, create_dl_layer_anchors, create_dl_layer_batch_normalization, create_dl_layer_box_proposals, create_dl_layer_box_targets, create_dl_layer_class_id_conversion, create_dl_layer_concat, create_dl_layer_convolution
• Matching (38/96): apply_deep_counting_model, clear_deformable_model, clear_descriptor_model, clear_ncc_model, clear_shape_model, create_deep_counting_model, deserialize_deformable_model, deserialize_descriptor_model, deserialize_ncc_model, deserialize_shape_model, get_deep_counting_model_param, get_deformable_model_contours
• File (0/53): close_file, copy_file, delete_file, deserialize_handle, deserialize_image, deserialize_object, deserialize_region, deserialize_tuple, deserialize_xld, file_exists, fnew_line, fread_bytes
• 3D Matching (15/59): add_deformable_surface_model_reference_point, add_deformable_surface_model_sample, apply_deep_matching_3d, apply_dl_model, clear_deformable_surface_matching_result, clear_deformable_surface_model, clear_shape_model_3d, clear_surface_matching_result, clear_surface_model, create_deep_matching_3d, deserialize_deformable_surface_model, deserialize_shape_model_3d
• Inspection (12/55): add_texture_inspection_model_image, clear_bead_inspection_model, clear_structured_light_model, clear_texture_inspection_model, clear_texture_inspection_result, clear_train_data_variation_model, clear_variation_model, close_ocv, deserialize_ocv, deserialize_structured_light_model, deserialize_texture_inspection_model, deserialize_variation_model
• Develop (0/37): dev_clear_obj, dev_clear_window, dev_close_inspect_ctrl, dev_close_tool, dev_close_window, dev_disp_text, dev_display, dev_error_var, dev_get_exception_data, dev_get_preferences, dev_get_system, dev_get_window

Version awareness (HALCON’s op set changes between releases)

Operator counts per scraped release: v12=2147, v13=2176, v2311=2381, v2411=2387, v2505=2411, v2605=2313 (union 2466). Coverage above is vs the primary scrape; the classification below is honest about which claimed Op.halcon names are stable vs release-specific.

• 927 claimed names exist in all scraped releases (stable).
• 54 version-drift (real, but only some releases): adapt_shape_model_high_noise (in 2311/2411/2505/2605); add_image_border (in 2311/2411/2505/2605); apply_texture_inspection_model (in 13/2311/2411/2505/2605); area_intersection_rectangle2 (in 2311/2411/2505/2605); bilateral_filter (in 13/2311/2411/2505/2605); convol_channels (in 2605); create_generic_shape_model (in 2311/2411/2505/2605); create_structured_light_model (in 2311/2411/2505/2605); create_texture_inspection_model (in 13/2311/2411/2505/2605); crop_rectangle2 (in 2311/2411/2505/2605); decode_structured_light_pattern (in 2311/2411/2505/2605); distance_cc_min_points (in 2311/2411/2505/2605); distance_point_line (in 2311/2411/2505/2605); distance_point_pluecker_line (in 2311/2411/2505/2605); dual_quat_compose (in 13/2311/2411/2505/2605); dual_quat_conjugate (in 13/2311/2411/2505/2605); dual_quat_interpolate (in 13/2311/2411/2505/2605); dual_quat_normalize (in 13/2311/2411/2505/2605); dual_quat_to_hom_mat3d (in 13/2311/2411/2505/2605); dual_quat_to_pose (in 13/2311/2411/2505/2605); dual_quat_to_screw (in 13/2311/2411/2505/2605); dual_quat_trans_line_3d (in 13/2311/2411/2505/2605); dual_quat_trans_point_3d (in 2311/2411/2505/2605); edges_object_model_3d (in 13/2311/2411/2505/2605); equ_histo_image_rect (in 2311/2411/2505/2605); find_box_3d (in 2311/2411/2505/2605); find_generic_shape_model (in 2311/2411/2505/2605); find_ncc_models (in 13/2311/2411/2505/2605); find_surface_model_image (in 13/2311/2411/2505/2605); fuse_object_model_3d (in 2311/2411/2505/2605); gen_canonical_variates_trans (in 2605); gen_image_warp_map (in 2505/2605); gen_savitzky_golay_filter (in 2605); gen_structured_light_pattern (in 2311/2411/2505/2605); guided_filter (in 13/2311/2411/2505/2605); height_width_ratio (in 2311/2411/2505/2605); height_width_ratio_xld (in 2311/2411/2505/2605); interleave_channels (in 13/2311/2411/2505/2605); intersection_region_contour_xld (in 2411/2505/2605); mean_image_shape (in 2311/2411/2505/2605); pluecker_line_to_point_direction (in 2311/2411/2505/2605); pluecker_line_to_points (in 2311/2411/2505/2605); point_direction_to_pluecker_line (in 2311/2411/2505/2605); point_pluecker_line_to_hom_mat3d (in 2311/2411/2505/2605); points_to_pluecker_line (in 2311/2411/2505/2605); pose_to_dual_quat (in 13/2311/2411/2505/2605); reconstruct_surface_structured_light (in 2311/2411/2505/2605); rectangularity_xld (in 2311/2411/2505/2605); refine_surface_model_pose_image (in 13/2311/2411/2505/2605); screw_to_dual_quat (in 13/2311/2411/2505/2605); segment_image_mser (in 13/2311/2411/2505/2605); test_region_points (in 2411/2505/2605); uncalibrated_photometric_stereo (in 2311/2411/2505/2605); watersheds_marker (in 2311/2411/2505/2605)
• 0 claimed names exist in no scraped release — genuine bad names / library-specific / voxel-3D, not version drift.

Honest reading

• HALCON’s ~2313 operators include large non-algorithmic chapters (Graphics / Tuple / System / File / Develop / Control) an algorithm-design engine does not target; real algorithmic headroom is smaller than the raw gap.
• Many operators are parametric variants of one family (collapse to fewer).
• Coverage counts a nearest analogue, not signature-level parity; per-operator typed stubs (data/halcon_stubs.json, with real Python signatures from the mvtec-halcon binding when available) track what is named vs implemented.

This site is open source. Improve this page.