nearle_daily/catalogue_backend
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

image repairment in own products

Browse Source
This commit is contained in:
sriram
2026-09-04 15:44:07 +05:30
parent 8059120ce3
commit bb30edaffa
4 changed files with 3054 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
app/services/data/produce_reference.json
View File

@@ -262,7 +262,7 @@
"wikipedia_title": "Chrysanthemum"
},
"Clam": {
"image_credit": "File:Am\u00eaijoas de Set\u00fabal.jpg",
"image_credit": "File:Amêijoas de Setúbal.jpg",
"image_note": "hand-picked: the article's lead image is a botanical plate",
"image_source": "wikimedia_commons",
"image_url": "https://thumb.wikimedia.org/wikipedia/commons/thumb/1/13/Am%C3%AAijoas_de_Set%C3%BAbal.jpg/960px-Am%C3%AAijoas_de_Set%C3%BAbal.jpg?utm_source=commons.wikimedia.org&utm_campaign=imageinfo&utm_content=thumbnail",
@@ -553,9 +553,10 @@
"wikipedia_title": "Ghee"
},
"Ginger": {
"image_credit": "Ginger",
"image_source": "wikipedia_lead",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/1/18/Koeh-146-no_text.jpg",
"image_credit": "Fresh ginger rhizome",
"image_note": "Wikipedia's lead image for Ginger is Koehler's Medizinal-Pflanzen plate 146 - a 19th-century botanical illustration of the whole plant, not the rhizome a shop sells. Hand-picked from Commons.",
"image_source": "wikimedia_commons",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/f/fb/Fresh_ginger_rhizome_01.jpg/960px-Fresh_ginger_rhizome_01.jpg",
"usda_description": "Ginger root, raw",
"usda_fdc_id": 169231,
"wikipedia_title": "Ginger"

1590
data/brand_image_repair_backup_20260904_151415.json Normal file
View File

File diff suppressed because it is too large Load Diff

1435
data/brand_image_repair_backup_20260904_152006.json Normal file
View File

File diff suppressed because it is too large Load Diff

26
scripts/repair_brand_images.py
View File

@@ -576,14 +576,36 @@ def repair(cur, conn, brands: List[str], apply: bool, timeout: int, workers: int
# three providers were caught serving a confidently wrong image.
# --recheck re-opens rows whose stored URL names neither the brand
# nor any distinctive word of the product, so they are re-searched
# under the corroboration rule rather than grandfathered in.
# under the corroboration rule rather than grandfathered in - and,
# for the produce bucket, any row that simply is not the reviewed
# image (see below).
have = {r["image_id"] for r in needs}
for row in rows:
if row["image_id"] in have:
continue
name = row.get("product_name") or ""
# A REVIEWED IMAGE OUTRANKS THE HEURISTIC.
#
# The corroboration test below asks whether the URL contains a
# distinctive word of the product, and it cannot tell an image OF
# the thing from an image merely NAMED after it. It passed, and so
# grandfathered in, `Bitter_Gourd_Curry.jpg` for Bitter Gourd,
# `Zunaid_Ahmed_Palak.jpg` - a photograph of a person - for Palak,
# and `BANANA_CREPE_CON_BASIL_LEAVES.jpg` for Basil Leaves. That
# left 57 rows untouched on the first full run.
#
# Where a person has already chosen the image, no heuristic gets a
# vote: if the stored URL is not the curated one, it is wrong.
if _is_own_products(suffix):
curated = produce_reference.image_url(name)
if curated and (row.get("image_url") or "") != curated:
needs.append(row)
continue
stored = _row_urls(row)
if stored and not any(
_names_product(u, row.get("product_name") or "", suffix) for u in stored
_names_product(u, name, suffix) for u in stored
):
needs.append(row)
totals["scanned"] += len(rows)