Datasets:
Tasks:
Document Question Answering
Size:
n<1K
Tags:
benchmark
document-ai
information-extraction
structured-extraction
key-information-extraction
ocr
License:
File size: 3,102 Bytes
8143f06 | 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 79 80 81 82 83 84 85 | #!/usr/bin/env python3
"""generate downscaled page images + docupipe extraction json for the landing-page benchmark modal."""
import json, os, re, subprocess, glob, tempfile
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
OUT = "/Users/urimerhav/Code/docupanda-website/public/benchmarks/docs"
MAX_PAGES = 5
WIDTH = 1000
html = open(os.path.join(ROOT, "docubench-explorer.html"), encoding="utf-8").read()
DATA = json.loads(re.search(r"const DATA = (\{.*?\});", html, re.S).group(1))
DOCS = {d["id"]: d for d in DATA["docs"]}
ids = json.load(open("/tmp/dbids.json"))
os.makedirs(OUT, exist_ok=True)
def sips_to_jpeg(src, dst):
subprocess.run(["sips", "-s", "format", "jpeg", "-s", "formatOptions", "72",
"-Z", str(WIDTH), src, "--out", dst],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
def render_pdf(path, outdir):
with tempfile.TemporaryDirectory() as td:
prefix = os.path.join(td, "pg")
subprocess.run(["pdftoppm", "-jpeg", "-r", "110", "-f", "1", "-l", str(MAX_PAGES), path, prefix],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
pages = sorted(glob.glob(prefix + "*"))
for i, p in enumerate(pages, 1):
sips_to_jpeg(p, os.path.join(outdir, f"p{i}.jpg"))
return len(pages)
def render_image(path, outdir):
sips_to_jpeg(path, os.path.join(outdir, "p1.jpg"))
return 1
def render_quicklook(path, outdir):
with tempfile.TemporaryDirectory() as td:
subprocess.run(["qlmanage", "-t", "-s", "1300", "-o", td, path],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
pngs = glob.glob(os.path.join(td, "*.png"))
if not pngs:
return 0
sips_to_jpeg(pngs[0], os.path.join(outdir, "p1.jpg"))
return 1
manifest = {}
for did in ids:
doc = DOCS[did]
ext = doc["ftype"]
src = glob.glob(os.path.join(ROOT, "documents", did + ".*"))
if not src:
print("MISSING doc", did); continue
src = src[0]
outdir = os.path.join(OUT, did)
os.makedirs(outdir, exist_ok=True)
if ext == "pdf":
n = render_pdf(src, outdir)
elif ext in ("jpeg", "jpg", "png", "tiff", "tif"):
n = render_image(src, outdir)
else: # csv, xlsx, docx, xml, txt, html
n = render_quicklook(src, outdir)
# docupipe high-effort extraction json -> standardization.json (just the data object)
res_path = os.path.join(ROOT, "results", "docupipe_high", did + ".json")
res = json.load(open(res_path, encoding="utf-8"))
json.dump(res.get("data", {}), open(os.path.join(outdir, "standardization.json"), "w", encoding="utf-8"),
ensure_ascii=False)
manifest[did] = {
"pages": n,
"ftype": ext,
"origPages": doc["pages"],
"score": round(res.get("score", 0), 4),
}
print(f"{did}: {ext:5s} pages={n} origPages={doc['pages']}")
json.dump(manifest, open(os.path.join(OUT, "manifest.json"), "w"), indent=0)
print("\nmanifest written:", len(manifest), "docs ->", OUT)
|