File size: 2,680 Bytes
c75b162 | 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 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 | """
Qualitative comparison: blurry input vs deblurred outputs.
No GT reference. Just visual quality comparison.
Columns: Blur input | EVSSM | VD-Diff | TURTLE (if available)
"""
import numpy as np
from pathlib import Path
from PIL import Image, ImageDraw
BASE = Path(__file__).parent.parent
METHODS = [
("blur", "Blur Input", BASE / "data/scannet_blur_proto/vddiff/test/blur"),
("evssm", "EVSSM", BASE / "data/evssm_deblurred"),
("vdiff", "VD-Diff", BASE / "data/vdiff_deblurred"),
("turtle", "TURTLE", BASE / "data/turtle_deblurred"),
]
def label_bar(text, w, h=36, bg=(30,30,30), fg=(255,255,255)):
img = Image.new("RGB", (w, h), bg)
draw = ImageDraw.Draw(img)
bbox = draw.textbbox((0,0), text)
tw, th = bbox[2]-bbox[0], bbox[3]-bbox[1]
draw.text(((w-tw)//2, (h-th)//2), text, fill=fg)
return np.array(img)
def load(path, scale=None):
img = Image.open(path).convert("RGB")
if scale and scale < 1.0:
w, h = img.size
img = img.resize((int(w*scale), int(h*scale)), Image.LANCZOS)
return np.array(img)
def make_row(scene, frame_idx, scale=0.45, crop=None):
fname = f"{frame_idx:06d}.png"
cols = []
for key, label, root in METHODS:
path = root / scene / fname
if not path.exists():
continue
img = load(path, scale=scale if crop is None else None)
if crop:
y1,x1,y2,x2 = crop
img = np.array(Image.open(path).convert("RGB"))[y1:y2, x1:x2]
bar = label_bar(label, img.shape[1])
cols.append(np.vstack([bar, img]))
if not cols:
return None
max_h = max(c.shape[0] for c in cols)
return np.hstack([np.vstack([c, np.zeros((max_h-c.shape[0], c.shape[1], 3), np.uint8)]) for c in cols])
def save_grid(rows, title, out_path):
title_bar = label_bar(title, rows[0].shape[1], h=30, bg=(60,30,30))
grid = np.vstack([title_bar] + rows)
Image.fromarray(grid).save(out_path, quality=93)
print(f" → {out_path}")
out_dir = BASE / "outputs/eval/qualitative_ngt"
out_dir.mkdir(parents=True, exist_ok=True)
scenes = ["scene0000_00", "scene0000_01", "scene0000_02"]
frame_idxs = [0, 15, 30, 45]
crops = {
"full": None,
"center": (280, 380, 620, 900), # y1,x1,y2,x2 — around centre
"top": (0, 300, 280, 900), # top region
}
for scene in scenes:
print(f"\n[{scene}]")
for crop_name, crop in crops.items():
rows = [r for fi in frame_idxs if (r := make_row(scene, fi, scale=0.42, crop=crop)) is not None]
if rows:
save_grid(rows, f"{scene} [{crop_name}]",
out_dir / f"{scene}_{crop_name}.jpg")
|