| |
| """Index the source library: one row per motion file in `Truebone_Z-OO/`. |
| |
| Step 1. Walks the original download in place, probes every BVH and FBX, and |
| writes `clips.csv`. Nothing is copied: `Truebone_Z-OO/` is the download exactly |
| as it shipped and stays the one copy of it, with `fbx/` and `animation/` |
| hardlinked to it further down the pipeline. Files that fail to parse are listed |
| in `excluded.csv` and left out. |
| |
| Skeleton and timeline come from the file headers, but `has_motion` does not: a |
| clip can carry ninety keyframes that all hold the same pose, so the curves are |
| decoded and their ranges measured. That is what makes this step take minutes |
| rather than seconds, and it is what keeps every T-pose out of `fbx/`. |
| |
| build_dataset.py -> normalize.py -> build_animation.py -> build_metadata.py |
| """ |
| import csv, os, re, sys, time |
| HERE = os.path.dirname(os.path.abspath(__file__)) |
| SCRIPTS = os.path.dirname(HERE) |
| ROOT = os.path.dirname(SCRIPTS) |
| sys.path[:0] = [HERE, os.path.join(SCRIPTS, "probes")] |
| from bvh_probe import probe as bvh_probe |
| from fbx_probe import probe as fbx_probe |
| from fbx_motion import probe as motion_probe |
|
|
| SRC_DIR = "Truebone_Z-OO" |
| SRC = sys.argv[1] if len(sys.argv) > 1 else os.path.join(ROOT, SRC_DIR) |
|
|
| CLIP_COLS=["file","group","format","clip_name","num_joints","frames","fps","duration_sec", |
| "root_joint","fbx_version","probe_error","has_motion"] |
|
|
| def main(): |
| groups=sorted(d for d in os.listdir(SRC) if os.path.isdir(os.path.join(SRC,d))) |
| rows=[]; excluded=[]; t0=time.time() |
| for g in groups: |
| gdir=os.path.join(SRC,g) |
| for dp,dn,fn in os.walk(gdir): |
| for f in sorted(fn): |
| ext=os.path.splitext(f)[1].lower() |
| if ext not in (".bvh",".fbx"): continue |
| p=os.path.join(dp,f) |
| rel=os.path.relpath(p,SRC) |
| r={"file":SRC_DIR+"/"+rel.replace(os.sep,"/"),"group":g, |
| "format":ext[1:],"clip_name":os.path.splitext(f)[0], |
| "num_joints":"","frames":"","fps":"","duration_sec":"", |
| "root_joint":"","fbx_version":"","probe_error":"","has_motion":""} |
| try: |
| if ext==".bvh": |
| b=bvh_probe(p,motion=True) |
| r.update(num_joints=b["num_joints"],frames=b["frames"],fps=b["fps"], |
| duration_sec=b["duration_sec"],root_joint=b["root"], |
| has_motion=str(b["has_motion"]).lower()) |
| else: |
| x=fbx_probe(p) |
| m=motion_probe(p) |
| r.update(num_joints=x["num_joints"],frames=x["keyframes"], |
| fps=x["fps"] or "",duration_sec=x["duration_sec"] or "", |
| root_joint=x["root_joint"],fbx_version=x["fbx_version"], |
| has_motion=str(not (m["static"] or m["no_curves"])).lower()) |
| except Exception as e: |
| r["probe_error"]=str(e)[:80] |
| excluded.append({"file":r["file"],"group":g,"reason":str(e)[:100]}); continue |
| if ext==".bvh" and int(r["frames"] or 0) < 2: |
| excluded.append({"file":r["file"],"group":g, |
| "reason":"truncated: %s frame(s)"%r["frames"]}); continue |
| rows.append(r) |
| print(" %-18s %d files %.0fs"%(g,sum(1 for x in rows if x["group"]==g),time.time()-t0),flush=True) |
|
|
| with open(os.path.join(ROOT,"clips.csv"),"w",newline="",encoding="utf-8") as fh: |
| w=csv.DictWriter(fh,fieldnames=CLIP_COLS); w.writeheader(); w.writerows(rows) |
| print("clips.csv ->",len(rows),"rows") |
|
|
| if excluded: |
| with open(os.path.join(ROOT,"excluded.csv"),"w",newline="",encoding="utf-8") as fh: |
| w=csv.DictWriter(fh,fieldnames=["file","group","reason"]); w.writeheader(); w.writerows(excluded) |
| print("excluded.csv ->",len(excluded),"rows (unparseable files skipped)") |
| print("%.0fs"%(time.time()-t0)) |
|
|
| if __name__=="__main__": main() |
|
|
