CheshireJCat/create-3d-model-skill / skills/create-3d-model/references/modules/reference-analysis-validator
reference-analysis-validator
VerifiedMeasure and validate supplied reference images, wireframes, texture atlases, and Blender renders before declaring a reconstruction 1:1. Use when an asset must match a template, when visual feedback says the output is off, when part counts must be exact, or before exporting a brand mascot/logo reconstruction. Pairs with reference-to-3d, contour-to-mesh, orthographic-registration, atlas-uv-fitting, and Blender MCP.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.2 KB |
| references/metrics-and-thresholds.md | markdown | 1.5 KB |
| scripts/reference_manifest_compiler.pyexecutable | script | 5.2 KB |
| scripts/render_overlay_validator.pyexecutable | script | 4.0 KB |
scripts/render_overlay_validator.py
View on GitHub#!/usr/bin/env python3
"""Compare reference/render images as masks and write validation metrics + overlay."""
from __future__ import annotations
import argparse, json
from pathlib import Path
import cv2
import numpy as np
try:
from skimage.metrics import structural_similarity as ssim
except Exception:
ssim = None
def mask(path, mode):
img=cv2.imread(str(path), cv2.IMREAD_UNCHANGED)
if img is None: raise RuntimeError(f'cannot read {path}')
if img.ndim==3 and img.shape[2]==4 and mode in ('auto','alpha'):
m=(img[:,:,3] > 8).astype('uint8')*255
rgb=cv2.cvtColor(img[:,:,:3], cv2.COLOR_BGR2RGB)
else:
rgb=cv2.cvtColor(img[:,:,:3], cv2.COLOR_BGR2RGB) if img.ndim==3 else cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
gray=cv2.cvtColor(rgb, cv2.COLOR_RGB2GRAY)
if mode=='edges': m=cv2.Canny(cv2.GaussianBlur(gray,(3,3),0),40,120)
elif mode=='dark_on_bright': _,m=cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
else: _,m=cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
return rgb, (m>0)
def bbox(m):
ys,xs=np.where(m)
if len(xs)==0: return None
return [int(xs.min()), int(ys.min()), int(xs.max()-xs.min()+1), int(ys.max()-ys.min()+1)]
def centroid(m):
ys,xs=np.where(m)
if len(xs)==0: return None
return [float(xs.mean()), float(ys.mean())]
def main():
ap=argparse.ArgumentParser()
ap.add_argument('--reference', required=True)
ap.add_argument('--render', required=True)
ap.add_argument('--out', required=True)
ap.add_argument('--overlay', default=None)
ap.add_argument('--mode', default='auto', choices=['auto','alpha','edges','bright_on_dark','dark_on_bright'], help='legacy: use same mask mode for reference and render')
ap.add_argument('--reference-mode', default=None, choices=['auto','alpha','edges','bright_on_dark','dark_on_bright'])
ap.add_argument('--render-mode', default=None, choices=['auto','alpha','edges','bright_on_dark','dark_on_bright'])
args=ap.parse_args()
reference_mode = args.reference_mode or args.mode
render_mode = args.render_mode or args.mode
ref_rgb, ref=mask(Path(args.reference), reference_mode)
ren_rgb, ren=mask(Path(args.render), render_mode)
if ref.shape != ren.shape:
ren=cv2.resize(ren.astype('uint8'), (ref.shape[1], ref.shape[0]), interpolation=cv2.INTER_NEAREST).astype(bool)
ren_rgb=cv2.resize(ren_rgb, (ref.shape[1], ref.shape[0]))
inter=np.logical_and(ref,ren).sum(); union=np.logical_or(ref,ren).sum()
iou=float(inter/union) if union else 0.0
mse=float(np.mean((ref.astype('float32')-ren.astype('float32'))**2))
ssim_val=None
if ssim:
ssim_val=float(ssim(ref.astype('uint8')*255, ren.astype('uint8')*255, data_range=255))
br, bn=bbox(ref), bbox(ren); cr, cn=centroid(ref), centroid(ren)
ref_ratio=float(ref.mean()); ren_ratio=float(ren.mean())
report={'iou':iou,'mse':mse,'ssim':ssim_val,'reference_mode':reference_mode,'render_mode':render_mode,'reference_foreground_ratio':ref_ratio,'render_foreground_ratio':ren_ratio,'reference_bbox':br,'render_bbox':bn,'reference_centroid':cr,'render_centroid':cn}
if max(ref_ratio, ren_ratio) > 0 and min(ref_ratio, ren_ratio) / max(ref_ratio, ren_ratio) < 0.25:
report['warning']='Mask foreground ratios differ strongly; compare same modality (silhouette-vs-silhouette or edge-vs-edge) before treating IoU as final.'
if cr and cn:
report['centroid_drift_px']=[cn[0]-cr[0], cn[1]-cr[1]]
out=Path(args.out); out.parent.mkdir(parents=True, exist_ok=True)
out.write_text(json.dumps(report, indent=2), encoding='utf-8')
if args.overlay:
overlay=np.zeros((*ref.shape,3), dtype=np.uint8)
overlay[ref]=[0,255,255] # cyan reference
overlay[ren]=[255,255,0] # yellow render
overlay[np.logical_and(ref,ren)]=[0,255,0]
overlay[np.logical_xor(ref,ren)]=[255,0,0]
cv2.imwrite(args.overlay, cv2.cvtColor(overlay, cv2.COLOR_RGB2BGR))
print(json.dumps(report, indent=2))
if __name__ == '__main__': main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 468k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |