File size: 1,017 Bytes
f9bc0c1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import json, re
from pathlib import Path
from PIL import Image, ImageDraw

root = Path.home() / "ocr"
repo = root / "OCR_INDRA"
data = root / "ocr-deva-ft/data/chunks-v4/test"
preds = repo / "results/stratified_1500/predictions_experiment1_all_models.jsonl"
out = repo / "results/stratified_1500/error_contact_sheet.png"

rows = [json.loads(s) for s in preds.open()]
strip = lambda s: re.sub(r"\s+", "", s)
bad = [r for r in rows if r["model"] == "Indra-FT"
       and not r["exact_match"]
       and strip(r["ground_truth"]) != strip(r["prediction"])][:10]

w, h = 900, 165
sheet = Image.new("RGB", (w, h * len(bad)), "white")
draw = ImageDraw.Draw(sheet)

for i, r in enumerate(bad):
    im = Image.open(data / r["image"]).convert("RGB")
    im.thumbnail((880, 110))
    y = i * h
    sheet.paste(im, (10, y))
    draw.text((10, y + 116), "REF: " + r["ground_truth"], fill="black")
    draw.text((10, y + 138), "FT:  " + r["prediction"], fill="blue")

sheet.save(out)
print("Saved:", out)
print("Crops:", len(bad))