File size: 4,347 Bytes
3e45900 | 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 | """Write final manifest (meta/), split each config into standalone <=300 MB ZIPs (classes never span parts)."""
import csv, json, os, zipfile
from collections import Counter, defaultdict
LIMIT = 300_000_000
os.makedirs('meta', exist_ok=True); os.makedirs('out', exist_ok=True)
TR = {'Blast': 'Blast', 'Promyelocyte': 'Promiyelosit', 'Myelocyte': 'Miyelosit', 'Metamyelocyte': 'Metamiyelosit',
'Band_neutrophil': 'Çomak nötrofil', 'Segmented_neutrophil': 'Segmentli nötrofil', 'Eosinophil': 'Eozinofil', 'Basophil': 'Bazofil',
'Lymphocyte': 'Lenfosit', 'Monocyte': 'Monosit', 'Proerythroblast': 'Proeritroblast', 'Erythroblast': 'Eritroblast',
'Plasma_cell': 'Plazma hücresi', 'Hairy_cell': 'Tüylü hücre (hairy cell)', 'Platelet': 'Trombosit', 'Artefact': 'Artefakt',
'Not_identifiable': 'Tanımlanamayan hücre', 'ALL': 'Akut lenfoblastik lösemi hücresi', 'Normal': 'Normal hücre',
'Parasitized': 'Plasmodium ile enfekte eritrosit', 'Uninfected': 'Enfekte olmayan eritrosit'}
SRC = {'bone_marrow_mll': ('Matek et al. 2021, Bone Marrow Cytomorphology (MLL / Helmholtz / Fraunhofer)', 'CC BY 4.0', 'https://doi.org/10.7937/TCIA.AXH3-T579'),
'pbc_acevedo': ('Acevedo et al. 2020, PBC dataset (Hospital Clinic of Barcelona)', 'CC BY 4.0', 'https://doi.org/10.17632/snkd93bnjr.1'),
'c_nmc_2019': ('Gupta & Gupta 2019, C-NMC 2019 (ISBI challenge)', 'CC BY 3.0', 'https://doi.org/10.7937/tcia.2019.dc64i46r'),
'nih_malaria': ('Rajaraman et al. 2018, NIH/NLM Malaria cell images', 'Public (U.S. NLM)', 'https://lhncbc.nlm.nih.gov/LHC-research/LHC-projects/image-processing/malaria-datasheet.html'),
'bccd': ('BCCD (Shenggan 2017), Roboflow 100 version', 'MIT (original) / CC BY 4.0 (RF100)', 'https://github.com/Shenggan/BCCD_Dataset')}
R = [r for r in csv.DictReader(open('manifest.csv')) if os.path.exists(os.path.join('dataset', r['file']))]
for r in R:
r['label_tr'] = TR.get(r['label'], '')
r['source_citation'], r['source_license'], r['source_url'] = SRC[r['source']]
fields = ['config', 'split', 'label', 'label_tr', 'file', 'source', 'original_label', 'original_file', 'patient_id', 'source_citation', 'source_license', 'source_url']
with open('meta/manifest.csv', 'w', newline='', encoding='utf-8-sig') as f:
w = csv.DictWriter(f, fieldnames=fields, extrasaction='ignore'); w.writeheader(); w.writerows(R)
stats = {}
for cfg in ('cls', 'all', 'malaria', 'det'):
rs = [r for r in R if r['config'] == cfg]
stats[cfg] = dict(images=len(rs), splits=dict(Counter(r['split'] for r in rs)),
classes={k: dict(Counter(r['split'] for r in rs if r['label'] == k)) for k in sorted({r['label'] for r in rs if r['label']})},
sources=dict(Counter(r['source'] for r in rs)), patients=len({r['patient_id'] for r in rs if r['patient_id']}))
json.dump(stats, open('meta/stats.json', 'w'), indent=1, ensure_ascii=False)
NAMES = {'cls': 'HEMATO-CLS', 'all': 'HEMATO-ALL', 'malaria': 'HEMATO-MALARIA', 'det': 'HEMATO-DET'}
for cfg in ('cls', 'all', 'malaria'):
by = defaultdict(list)
for r in R:
if r['config'] == cfg: by[r['label']].append(r['file'])
# every part must contain several classes (a single-class ZIP is imported as unlabeled "flat" data),
# so files are dealt round-robin across parts: each part gets a slice of every class.
allf = [f for lab in sorted(by) for f in sorted(by[lab])]
total = sum(os.path.getsize(os.path.join('dataset', f)) for f in allf)
n = max(1, -(-total // LIMIT))
parts = [allf[i::n] for i in range(n)]
for i, files in enumerate(parts, 1):
zn = f'out/{NAMES[cfg]}_part{i:02d}-of-{len(parts):02d}.zip'
with zipfile.ZipFile(zn, 'w', zipfile.ZIP_STORED) as z:
for f in files: z.write(os.path.join('dataset', f), f.split('/', 1)[1]) # drop the config prefix
print(zn, round(os.path.getsize(zn) / 1e6), 'MB', len(files))
with zipfile.ZipFile('out/HEMATO-DET.zip', 'w', zipfile.ZIP_DEFLATED) as z:
for root, _, files in os.walk('dataset/det'):
for fn in files:
p = os.path.join(root, fn); z.write(p, os.path.relpath(p, 'dataset/det'))
print('out/HEMATO-DET.zip', round(os.path.getsize('out/HEMATO-DET.zip') / 1e6), 'MB')
print(json.dumps({k: (v['images'], v['splits']) for k, v in stats.items()}))
|