File size: 1,736 Bytes
876f3c3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | #!/usr/bin/env python3
"""Validate replayed RoboTwin GT RGB against the original RGB HDF5."""
import argparse
from pathlib import Path
import cv2
import h5py
import numpy as np
def decode(x):
return cv2.imdecode(np.frombuffer(x, np.uint8), cv2.IMREAD_COLOR)
def main():
p = argparse.ArgumentParser()
p.add_argument("--original", type=Path, required=True)
p.add_argument("--ground-truth", type=Path, required=True)
p.add_argument("--camera", default="head_camera")
p.add_argument("--samples", type=int, default=8)
args = p.parse_args()
rows = []
for src in sorted(args.original.glob("episode*.hdf5")):
dst = args.ground_truth / src.name
if not dst.exists():
continue
with h5py.File(src, "r") as a, h5py.File(dst, "r") as b:
x = a[f"observation/{args.camera}/rgb"]
y = b[f"observation/{args.camera}/rgb"]
if len(x) != len(y):
rows.append((src.stem, len(x), len(y), np.inf))
continue
ids = np.linspace(0, len(x) - 1, min(args.samples, len(x)), dtype=int)
errors = []
for i in ids:
u, v = decode(x[i]), decode(y[i])
errors.append(float(np.mean(np.abs(u.astype(np.float32) - v.astype(np.float32)))))
rows.append((src.stem, len(x), len(y), float(np.mean(errors))))
if not rows:
raise SystemExit("no comparable episodes")
for row in rows:
print("episode=%s original=%d replay=%d mean_abs=%.4f" % row)
bad = [r for r in rows if not np.isfinite(r[3]) or r[3] > 5.0]
print(f"checked={len(rows)} bad={len(bad)}")
raise SystemExit(1 if bad else 0)
if __name__ == "__main__":
main()
|