code wiki / (root) / nx_image_grade_v2.nx

nx_image_grade_v2.nx

buildroot/runtime/nx_image_grade_v2.nx

20419 B543 linesdepth 8pulls 22 transitivereach 3 importersview sourcekind tooltopic image
docsdependenciesstructsconstsfunctions

about

nx_image_grade_v2.nx -- landmark-aware image grading composer. CAPABILITY_COMPLETENESS: PARTIAL MISSING_CAPABILITIES: - face keypoint detection (eye/nose/mouth within head bbox): blocks nx_micro_expression's full invocation; v2 returns head BBOX but doesn't yet sub-detect facial features - hand keypoint detection / finger segmentation: blocks nx_finger_count's full invocation - pose joint-angle extraction: blocks nx_pose_naturalness from running end-to-end - VLM grader source (triangulation second-source) COVERED (PROMOTED FROM v1): - Segmenter + landmark labeler composed - nx_silhouette_integrity invoked on torso blob -> SHARP / SOFT / MELTED verdict - nx_limb_count invoked with landmark output -> OK / EXTRA / FLOATING_LIMB / MISSING / etc. - Skin-localized Lab + ITA + undertone (same as v1) - Skin coverage Q10 (same as v1) - 6 anatomical-region bboxes exposed in report for caller to feed downstream tier-axes Composes: nx_png_decoder + nx_jpeg_decoder (format dispatch) nx_segmenter_classical (skin connected-components) nx_landmark_anatomy (head/torso/arm/leg labeling) nx_silhouette_integrity (melty detector on torso) nx_limb_count (anatomy coherence) nx_lab_from_rgb + nx_skin_tone_ita + nx_undertone_class nx_verdict_phrase genealogy_id: substrate_image_grade_v2_landmark_aware_2026_05_16 lineage_id: nx_image_grade_v2

dependencies 13 imports · 2 importers

nx_syscalls.nx nx_runtime.nx nx_tier.nx nx_png_decoder.nx nx_jpeg_decoder.nx nx_verdict_phrase.nx nx_lab_from_rgb.nx nx_skin_tone_ita.nx nx_undertone_class.nx nx_segmenter_classical.nx nx_image_grade_v2.nx nx_actor_role_image_grader.nx nx_image_grade_cli.nx

diagram shows first 10 each side; +3 more imports, +0 more importers in the complete lists below.

imports: nx_syscalls.nxnx_runtime.nxnx_tier.nxnx_png_decoder.nxnx_jpeg_decoder.nxnx_verdict_phrase.nxnx_lab_from_rgb.nxnx_skin_tone_ita.nxnx_undertone_class.nxnx_segmenter_classical.nxnx_landmark_anatomy.nxnx_silhouette_integrity.nxnx_limb_count.nx

imported by: nx_actor_role_image_grader.nxnx_image_grade_cli.nx

call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown

main sys_mmap nx_image_grade_v2 sys_mmap ↻ _igv2_detect_format nx_png_decode sys_mmap ↻ _png_check_signature _png_read_u32_be nx_crc32 nx_crc32_update nx_crc32_init nx_crc32_table_at _png_n_channels _png_bytes_per_pixel _png_n_channels ↻ _png_a7_expected _png_a7_col0 _png_a7_coli _png_a7_row0 _png_a7_rowi nx_zlib_inflate sys_mmap ↻ nx_deflate_inflate sys_mmap ↻ nx_bitstream_alloc sys_mmap ↻ _deflate_inflate_block nx_bitstream_read_lsb nx_bitstream_byte_align nx_bitstream_read_byte_ali _deflate_build_static_litl _deflate_build_static_dist _deflate_decode_huffman_bl sys_mmap ↻ _deflate_build_dynamic_tre _deflate_bytes_consumed adler32 _png_a7_deinterlace _png_a7_col0 ↻

structs

76struct NxIgv2AxisVerdict
89struct NxIgv2Report

consts

50const NX_MAGIC_1024: i64 = 1024
54const NX_IGV2_FORMAT_UNKNOWN: nx_int = 0
55const NX_IGV2_FORMAT_PNG: nx_int = 1
56const NX_IGV2_FORMAT_JPEG: nx_int = 2
58const NX_IGV2_OK: nx_int = 0
59const NX_IGV2_ERR_TOO_SHORT: nx_int = 1
60const NX_IGV2_ERR_UNKNOWN_FORMAT: nx_int = 2
61const NX_IGV2_ERR_DECODER: nx_int = 3
62const NX_IGV2_ERR_NULL_PIXELS: nx_int = 4
66const NX_IGV2_AXIS_OK: nx_int = 0
67const NX_IGV2_AXIS_PENDING_KEYPOINT: nx_int = 1 // need face/hand keypoint detection
68const NX_IGV2_AXIS_PENDING_POSE_ESTIMATE: nx_int = 2 // need joint-angle extractor
69const NX_IGV2_AXIS_PENDING_VLM: nx_int = 3
70const NX_IGV2_AXIS_NOT_APPLICABLE: nx_int = 4
71const NX_IGV2_AXIS_NO_SKIN_DETECTED: nx_int = 5
72const NX_IGV2_AXIS_INSUFFICIENT_LANDMARKS: nx_int = 6
85const NX_IGV2_AXIS_VERDICT_BYTES: nx_size = 48
108const NX_IGV2_REPORT_BYTES: nx_size = 112
111const NX_IGV2_AX_DIMENSIONS: nx_int = 0
112const NX_IGV2_AX_LUMINANCE: nx_int = 1
113const NX_IGV2_AX_SKIN_TONE_ITA: nx_int = 4
114const NX_IGV2_AX_UNDERTONE: nx_int = 12
115const NX_IGV2_AX_SILHOUETTE: nx_int = 28
116const NX_IGV2_AX_VEIN_SIGNAL: nx_int = 16
117const NX_IGV2_AX_SUBSURFACE: nx_int = 17
118const NX_IGV2_AX_MICRO_EXPRESSION: nx_int = 22
119const NX_IGV2_AX_LIMB_COUNT: nx_int = 25
120const NX_IGV2_AX_POSE_NATURALNESS: nx_int = 26
121const NX_IGV2_AX_FINGER_COUNT: nx_int = 24
122const NX_IGV2_AX_SHADOW_DIRECTION: nx_int = 23
123const NX_IGV2_AX_SURFACE_PLANE: nx_int = 27
124const NX_IGV2_AX_SKIN_COVERAGE: nx_int = 100
125const NX_IGV2_AX_LANDMARK_CONF: nx_int = 101
127const NX_IGV2_N_AXES: nx_int = 15

functions

131func _igv2_detect_format(buf: *u8, n: nx_int) -> nx_int
called by 2: nx_image_grade_v2main
152func _igv2_is_skin_pixel(r: nx_int, g: nx_int, b: nx_int) -> nx_int
called by 2: nx_image_grade_v2main
173func _igv2_set_axis(axes: *NxIgv2AxisVerdict, idx: nx_int,
called by 1: nx_image_grade_v2 calls 2: sys_mmapnx_verdict_phrase_for
223func nx_image_grade_v2(input: *u8, input_size: nx_int) -> *NxIgv2Report
called by 3: _im_step_grademainmain calls 14: sys_mmap_igv2_detect_formatnx_png_decodenx_jpeg_decode_igv2_is_skin_pixelnx_segmenter_connected+8
508func main() -> nx_int
calls 4: sys_mmapnx_image_grade_v2_igv2_detect_format_igv2_is_skin_pixel