Skip to content

MAX-786/claude-3d-harness / library/cc/texture-driven-mesh-fitting

texture-driven-mesh-fitting

Verified

Reshape source-locked mesh boundaries so the geometry fits texture-atlas or decal contours 1:1 before final UV/material work. Use when textures look off because the model behind them does not match the source texture region, or when the mesh must adapt to the texture instead of stretching the texture onto an approximate mesh.

FileKindSize
SKILL.mdskill md2.9 KB
scripts/contour_correspondence_report.pyscript2.3 KB

scripts/contour_correspondence_report.py

View on GitHub
#!/usr/bin/env python3
from __future__ import annotations
import argparse,json,cv2,numpy as np
from pathlib import Path

def mask(path):
    img=cv2.imread(str(path),cv2.IMREAD_UNCHANGED)
    if img is None: raise RuntimeError(path)
    if img.ndim==3 and img.shape[2]==4: m=img[:,:,3]>8
    else:
        gray=cv2.cvtColor(img[:,:,:3],cv2.COLOR_BGR2GRAY) if img.ndim==3 else img
        _,m=cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU); m=m>0
    return m.astype('uint8')

def largest_contour(m):
    cs,_=cv2.findContours(m,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE)
    if not cs: return None
    return max(cs,key=cv2.contourArea)[:,0,:].astype('float32')

def bbox(c):
    xs=c[:,0]; ys=c[:,1]
    return [float(xs.min()),float(ys.min()),float(xs.max()-xs.min()+1),float(ys.max()-ys.min()+1)]

def sample(c,n):
    pts=c
    d=np.sqrt(((np.roll(pts,-1,axis=0)-pts)**2).sum(axis=1))
    s=np.r_[0,np.cumsum(d)]
    total=s[-1]
    out=[]
    for target in np.linspace(0,total,n,endpoint=False):
        i=np.searchsorted(s,target,side='right')-1
        t=(target-s[i])/max(d[i],1e-9)
        out.append((pts[i]*(1-t)+pts[(i+1)%len(pts)]*t).tolist())
    return np.array(out,dtype='float32')

def main():
    ap=argparse.ArgumentParser(); ap.add_argument('--source-mask',required=True); ap.add_argument('--product-mask',required=True); ap.add_argument('--out',required=True); ap.add_argument('--samples',type=int,default=64)
    args=ap.parse_args()
    sc=largest_contour(mask(args.source_mask)); pc=largest_contour(mask(args.product_mask))
    if sc is None or pc is None: raise SystemExit('missing contour')
    ss=sample(sc,args.samples); ps=sample(pc,args.samples); delta=ps-ss
    rep={'schema':'contour_correspondence_report.v1','source_mask':args.source_mask,'product_mask':args.product_mask,'samples':args.samples,'source_bbox':bbox(sc),'product_bbox':bbox(pc),'mean_delta_px':[float(x) for x in delta.mean(axis=0)],'max_sample_error_px':float(np.sqrt((delta*delta).sum(axis=1)).max()),'sample_deltas_px':delta.tolist()}
    Path(args.out).parent.mkdir(parents=True,exist_ok=True); Path(args.out).write_text(json.dumps(rep,indent=2),encoding='utf-8')
    print(json.dumps({'mean_delta_px':rep['mean_delta_px'],'max_sample_error_px':rep['max_sample_error_px']},indent=2))
if __name__=='__main__': main()
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall