mirror of
https://github.com/ultralytics/ultralytics
synced 2026-05-10 09:04:26 +00:00
ebaebcb51d
Parses COCO AP logs from original DEIMv2 repo, a local replication run, and Ultralytics training to produce a side-by-side mAP curve visualization.
79 lines
2.7 KiB
Python
79 lines
2.7 KiB
Python
import re
|
|
from pathlib import Path
|
|
import matplotlib
|
|
matplotlib.use("Agg")
|
|
import matplotlib.pyplot as plt
|
|
|
|
HERE = Path(__file__).parent
|
|
OUT_PATH = HERE / "deimv2_dinov3_l_coco_map.png"
|
|
|
|
# --- Parsers ---
|
|
deim_pattern = re.compile(
|
|
r"Average Precision\s+\(AP\) @\[ IoU=0\.50:0\.95 \| area= all \| maxDets=100 \] = ([0-9.]+)"
|
|
)
|
|
ult_pattern = re.compile(
|
|
r"^\s+all\s+\d+\s+\d+\s+[\d.]+\s+[\d.]+\s+[\d.]+\s+([\d.]+)\s*$"
|
|
)
|
|
|
|
def parse_map(log_path, pattern):
|
|
values = []
|
|
with open(log_path, "r") as f:
|
|
for line in f:
|
|
m = pattern.search(line)
|
|
if m:
|
|
values.append(float(m.group(1)))
|
|
return values
|
|
|
|
deim_map = parse_map(HERE / "deimv2_dinov3_l_coco.log", deim_pattern)
|
|
repli_map = parse_map(HERE / "deimv2_large_bs16.log", deim_pattern)
|
|
ult_map = parse_map(HERE / "output.log", ult_pattern)
|
|
|
|
for label, vals in [("DEIM (original)", deim_map), ("DEIM (replication)", repli_map), ("Ultralytics", ult_map)]:
|
|
best = max(vals)
|
|
print(f"{label}: {len(vals)} epochs, best={best:.4f} @ ep{vals.index(best)}, final={vals[-1]:.4f}")
|
|
|
|
# --- Plot ---
|
|
fig, ax = plt.subplots(figsize=(12, 6))
|
|
|
|
COLORS = {"deim": "#1f77b4", "repli": "#2ca02c", "ult": "#ff7f0e"}
|
|
|
|
OFFSET = 0.003
|
|
ult_map_shifted = [v + OFFSET for v in ult_map]
|
|
|
|
series = [
|
|
(deim_map, COLORS["deim"], "-", "Original DEIMv2 repo (deimv2_dinov3_l_coco.log)"),
|
|
(repli_map, COLORS["repli"], "-", "DEIMv2 repo replication (deimv2_large_bs16.log)"),
|
|
(ult_map, COLORS["ult"], "-", "Ultralytics training (output.log)"),
|
|
(ult_map_shifted, COLORS["ult"], "--", f"Ultralytics training +{OFFSET:.3f} offset (output.log)"),
|
|
]
|
|
|
|
for vals, color, ls, label in series:
|
|
epochs = list(range(len(vals)))
|
|
best_ep = vals.index(max(vals))
|
|
best_v = max(vals)
|
|
|
|
ax.plot(epochs, vals, color=color, linewidth=2, linestyle=ls, label=label,
|
|
alpha=0.6 if ls == "--" else 1.0)
|
|
ax.scatter([best_ep], [best_v], color=color, s=60, zorder=5,
|
|
marker="^" if ls == "--" else "o")
|
|
if ls != "--":
|
|
ax.annotate(
|
|
f"Best: {best_v:.3f} @ ep{best_ep}",
|
|
xy=(best_ep, best_v),
|
|
xytext=(best_ep + 1.5, best_v + 0.0035),
|
|
fontsize=9,
|
|
color=color,
|
|
arrowprops=dict(arrowstyle="->", color=color, lw=1.2),
|
|
)
|
|
|
|
ax.set_xlabel("Epoch", fontsize=12)
|
|
ax.set_ylabel("mAP (COCO, IoU 0.50:0.95)", fontsize=12)
|
|
ax.set_title("DEIMv2 DINOv3-L COCO Training — mAP per Epoch", fontsize=13, fontweight="bold")
|
|
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda y, _: f"{y:.3f}"))
|
|
ax.grid(True, linestyle="--", alpha=0.4)
|
|
ax.legend(loc="lower right", fontsize=9)
|
|
|
|
plt.tight_layout()
|
|
plt.savefig(OUT_PATH, dpi=150)
|
|
print(f"Saved → {OUT_PATH}")
|