hkust-aaron's picture
Add annotated sample pack: quality, timeline, clip captions, keyframe masks (v1/v2), SAM 3.1 dense tracks
07f5340 verified
Raw History Blame Contribute Delete
2.77 kB
"""Per-frame quality metrics for the whole video + review contact sheets.
Usage: python 01_frame_metrics.py <video> <out_csv> <sheet_dir>
Metrics (every source frame):
sharpness variance of Laplacian on the grey image (higher = sharper)
brightness mean grey level 0-255
red_ratio mean R / (R+G+B); intracorporeal tissue is strongly red
green_ratio mean G / (R+G+B)
saturated fraction of pixels with grey >= 250 (glare / light flooding)
dark fraction of pixels with grey <= 15
motion mean absolute grey difference to the previous frame
"""
import csv
import sys
from pathlib import Path
import cv2
import numpy as np
video, out_csv, sheet_dir = sys.argv[1], sys.argv[2], Path(sys.argv[3])
sheet_dir.mkdir(parents=True, exist_ok=True)
cap = cv2.VideoCapture(video)
fps = cap.get(cv2.CAP_PROP_FPS)
SHEET_STEP_S = 4 # one thumbnail every 4 s for the overview sheets
COLS, ROWS, TW, TH = 5, 6, 288, 230
thumbs = []
prev = None
rows = []
idx = 0
while True:
ok, frame = cap.read()
if not ok:
break
grey = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
small = cv2.resize(grey, (180, 144), interpolation=cv2.INTER_AREA)
b, g, r = [c.astype(np.float64).mean() for c in cv2.split(frame)]
total = r + g + b + 1e-9
motion = 0.0 if prev is None else float(np.abs(small.astype(np.int16) - prev).mean())
prev = small.astype(np.int16)
rows.append([
idx, round(idx / fps, 3),
round(float(cv2.Laplacian(grey, cv2.CV_64F).var()), 2),
round(float(grey.mean()), 2),
round(r / total, 4), round(g / total, 4),
round(float((grey >= 250).mean()), 4),
round(float((grey <= 15).mean()), 4),
round(motion, 3),
])
if idx % int(round(fps * SHEET_STEP_S)) == 0:
t = cv2.resize(frame, (TW, TH), interpolation=cv2.INTER_AREA)
cv2.rectangle(t, (0, 0), (TW, 16), (0, 0, 0), -1)
cv2.putText(t, f"{idx / fps:6.1f}s f{idx}", (4, 12), cv2.FONT_HERSHEY_SIMPLEX, 0.4, (255, 255, 255), 1)
thumbs.append(t)
idx += 1
cap.release()
with open(out_csv, "w", newline="") as f:
w = csv.writer(f)
w.writerow(["frame_index", "timestamp_seconds", "sharpness", "brightness", "red_ratio",
"green_ratio", "saturated", "dark", "motion"])
w.writerows(rows)
per = COLS * ROWS
for s in range(0, len(thumbs), per):
chunk = thumbs[s:s + per]
sheet = np.zeros((ROWS * TH, COLS * TW, 3), np.uint8)
for i, t in enumerate(chunk):
y, x = divmod(i, COLS)
sheet[y * TH:(y + 1) * TH, x * TW:(x + 1) * TW] = t
cv2.imwrite(str(sheet_dir / f"overview_{s // per:02d}.jpg"), sheet, [cv2.IMWRITE_JPEG_QUALITY, 85])
print("frames", idx, "fps", fps, "sheets", (len(thumbs) + per - 1) // per)