Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel
"""Compact source-only research summaries; outputs are ignored local evidence."""
import hashlib
import json
import runpy
import struct
from pathlib import Path
import numpy as np
ROOT = Path(__file__).resolve().parents[1]
OUT = ROOT / 'artifacts/local/humanoid-recon-x0'
def names(array):
return array.tobytes().decode('utf-8').rstrip('\0').split('\0')
def main():
runpy.run_path(str(ROOT / 'scripts/fetch-humanoid-recon-x0.py'))['verify_inputs']()
inspection = json.loads((OUT / 'inspection.json').read_text())
summary = {}
for name, data in inspection['candidates'].items():
folder = OUT / 'CharMorph-db/characters' / name
item = dict(mesh=data['meshes'][0], rigs=[])
item['mesh'].pop('vertex_groups')
item['mesh'].pop('shape_keys')
for rig in data['rigs']:
item['rigs'].append(dict(name=rig['name'], count=len(rig['bones']),
roots=[b['name'] for b in rig['bones'] if b['parent'] is None],
sample_bones=[b for b in rig['bones'] if b['name'] in
('upper_arm.L', 'forearm.L', 'hand.L', 'thigh.L', 'shin.L', 'foot.L', 'toe.L', 'spine', 'spine.006')]))
item['morphs'] = {}
for level in ('L2', 'L3'):
pack = folder / ('morphs/' + level + '.npz')
if pack.exists():
with np.load(pack, allow_pickle=False) as z:
n = names(z['names'])
item['morphs'][level] = dict(count=len(n), examples=n[:12], storage='packed')
else:
files = sorted((folder / 'morphs' / level).glob('*.np*'))
n = [p.stem for p in files]
item['morphs'][level] = dict(count=len(n), examples=n[:12], storage='individual dense/sparse',
dense_count=sum(p.suffix == '.npy' for p in files), sparse_count=sum(p.suffix == '.npz' for p in files))
item['groups'] = {}
for relative in ('weights/rigify.npz', 'joints/rigify.npz'):
with np.load(folder / relative, allow_pickle=False) as z:
n = names(z['names'])
sums = np.zeros(item['mesh']['vertices'])
np.add.at(sums, z['idx'], z['weights'])
item['groups'][relative] = dict(count=len(n), entries=len(z['idx']),
index_min=int(z['idx'].min()), index_max=int(z['idx'].max()),
weights_finite=bool(np.isfinite(z['weights']).all()),
weight_min=float(z['weights'].min()), weight_max=float(z['weights'].max()),
per_vertex_sum_min=float(sums.min()), per_vertex_sum_max=float(sums.max()),
unweighted_vertices=int((sums == 0).sum()), examples=n[:10])
summary[name] = item
(OUT / 'candidate-summary.json').write_text(json.dumps(summary, indent=2), encoding='utf-8')
samples = []
for file in sorted((OUT / 'vrm-specification/samples').rglob('*.vrm')):
raw = file.read_bytes()
magic, version, length = struct.unpack_from('<4sII', raw)
assert magic == b'glTF' and version == 2 and length == len(raw)
size, kind = struct.unpack_from('<II', raw, 12)
assert kind == 0x4E4F534A
gltf = json.loads(raw[20:20 + size])
vrm = gltf.get('extensions', {}).get('VRMC_vrm')
if not vrm:
continue
springs = gltf.get('extensions', {}).get('VRMC_springBone', {})
samples.append(dict(path=file.relative_to(OUT / 'vrm-specification').as_posix(),
sha256=hashlib.sha256(raw).hexdigest(), generator=gltf.get('asset', {}).get('generator'),
meta=vrm['meta'], version=vrm['specVersion'], extensions=gltf.get('extensionsUsed', []),
human_bones=len(vrm['humanoid']['humanBones']), meshes=len(gltf.get('meshes', [])),
primitives=sum(len(m['primitives']) for m in gltf.get('meshes', [])),
primitive_morph_target_counts=[len(p.get('targets', [])) for m in gltf.get('meshes', []) for p in m['primitives']],
skins=len(gltf.get('skins', [])), skin_joint_counts=[len(s['joints']) for s in gltf.get('skins', [])],
materials=len(gltf.get('materials', [])),
expressions={k:list(v) for k,v in vrm.get('expressions', {}).items()},
look_at=vrm.get('lookAt'), first_person=vrm.get('firstPerson'),
springs=len(springs.get('springs', [])), colliders=len(springs.get('colliders', [])),
collider_groups=len(springs.get('colliderGroups', [])),
constrained_nodes=sum('VRMC_node_constraint' in n.get('extensions', {}) for n in gltf.get('nodes', []))))
(OUT / 'vrm-summary.json').write_text(json.dumps(samples, indent=2), encoding='utf-8')
if __name__ == '__main__':
main()