robometer/scripts/plot_one_episode.py

#!/usr/bin/env python3
"""Plot the 5 prefix-mode progress curves for a single episode.

Usage:
  python plot_one_episode.py chunk-000_episode_000039
  python plot_one_episode.py chunk-000_episode_000039 --out /tmp/x.png
"""
import argparse
import json
from pathlib import Path

import numpy as np
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt

MODES = ["uniform", "front_biased", "back_biased", "random_seed0", "random_seed1"]
COLORS = {"uniform": "tab:blue", "front_biased": "tab:orange",
          "back_biased": "tab:green", "random_seed0": "tab:red",
          "random_seed1": "tab:purple"}
FRACS = ["1/4", "2/4", "3/4", "end"]

p = argparse.ArgumentParser()
p.add_argument("episode_dir", help="e.g. chunk-000_episode_000039")
p.add_argument("--results-root",
               default=str(Path(__file__).resolve().parent.parent / "results_full"))
p.add_argument("--out", default=None)
a = p.parse_args()

ep_dir = Path(a.results_root) / "episode_results" / a.episode_dir
out = Path(a.out) if a.out else ep_dir / "curves.png"

data = {}
for m in MODES:
    f = ep_dir / f"{m}.json"
    if f.exists():
        data[m] = json.loads(f.read_text())
if not data:
    raise SystemExit(f"no mode json found in {ep_dir}")

meta0 = next(iter(data.values()))
n = meta0["pool_n"]
cps = [int((n - 1) * k / 4) for k in (1, 2, 3, 4)]
xs = np.arange(n)  # pool frame index

fig, ax = plt.subplots(figsize=(13, 6))
for m in MODES:
    if m not in data:
        continue
    ax.plot(xs, data[m]["scores_100"], color=COLORS[m], lw=1.6, label=m)
for frac, t in zip(FRACS, cps):
    ax.axvline(t, color="0.6", ls=":", lw=1)
    ax.text(t, 96, f"{frac} (t={t})", ha="center", fontsize=9, color="0.4")

# secondary axis: original video frame index (pool is downsampled)
raw_per_pool = meta0["total_raw_frames"] / max(n, 1)
sec_ax = ax.secondary_xaxis(
    "top",
    functions=(lambda i: i * raw_per_pool, lambda r: r / raw_per_pool))
sec_ax.set_xlabel(f"original video frame index "
                  f"(raw {meta0['total_raw_frames']} frames @ {meta0['native_fps']:.0f} fps)")

scores = {frac: [data[m]["scores_100"][t] for m in MODES if m in data]
          for frac, t in zip(FRACS, cps)}
rngs = ", ".join(f"{frac}: {max(v)-min(v):.1f}" for frac, v in scores.items())
meta = next(iter(data.values()))
ax.set_title(f"{a.episode_dir}  ({meta['camera']}, pool={n})\n"
             f"task: {meta['task'][:100]}\nPrefix Range @ checkpoints -> {rngs}",
             fontsize=10)
ax.set_xlabel("pool frame index (3fps-downsampled sequence)")
ax.set_xlim(0, n - 1)
ax.set_ylabel("progress score (0-100)")
ax.set_ylim(0, 100)
ax.grid(alpha=0.25)
ax.legend(fontsize=9)
fig.tight_layout()
fig.savefig(out, dpi=140)
print("saved:", out)